1. Introduction
The El Niño–Southern Oscillation (ENSO) phenomenon, known as the most prominent natural interannual climate signal, has widespread effects on the global climate system and the ecological systems of the tropical Pacific. Any strong change in ENSO characteristics can have serious climatic and ecological consequences (Latif and Keenlyside 2009).
In the last two decades, many studies have reported prominent changes in the frequency of ENSO during global warming. However, whether they are closely related to each other remains to be studied further. For example, Trenberth and Hoar (1996) pointed out there has been a tendency toward more frequent El Niño events and fewer La Niña events since the late 1970s and that the prolonged 1990–95 El Niño event was highly unusual, which may serve as evidence of the effect of anthropogenically forced global warming and related climate change. However, this conclusion was challenged by two subsequent studies (Harrison and Larkin 1997; Rajagopalan et al. 1997), which suggested that the 1990–95 El Niño event might have been part of the natural variability. However, using the definition of El Niño proposed by Trenberth (1997), Trenberth and Hoar (1997) argued that the (then) newly underway El Niño event of 1997 further confirmed that the tendency for more El Niño and fewer La Niña events since the late 1970s is very unlikely to be accounted for solely by natural variability. Latif et al. (1997) argued that there appears to be a decadal mode of tropical sea surface temperature (SST) variability that is independent of ENSO, and that this decadal mode dominated the anomalous 1990s. Gu and Philander (1997) argued that the anomalous early 1990s could be attributed to an interdecadal climate fluctuation that involves changes in the properties of the equatorial thermocline arising as a result of an influx of water with anomalous temperatures from higher latitudes. Some studies suggested that judging whether properties of ENSO are changing depends on how we view the background climate state (Fedorov and Philander 2000, 2001; Philander and Fedorov 2003): if the interdecadal variation is chosen as the reference temperature (background climate state), then the apparently prolonged El Niño that started in 1992 could be alternatively interpreted as the persistence of warm background conditions (Fedorov and Philander 2000, 2001). They also pointed out the limitation of arbitrary choice of the time-averaged temperature of the past century as a reference line for ENSO definition because such a selected reference temperature is an aspect of the background climate state which, in reality, is changing continually (Fedorov and Philander 2001). On the basis of an analysis of a reconstructed Niño-3.4 (5°S–5°N, 170°–120°W) SST index from high-resolution fossil coral records, Cobb et al. (2003) reported that a broad range of ENSO behavior, when ENSO is defined as interannual variability (2–7 yr) only, correlates poorly with the mean climate conditions over the last millennium. They suggested that most of the ENSO variability might have arisen from dynamics internal to the ENSO system itself. A recent study suggests that the frequency of El Niño events is related to the Atlantic multidecadal oscillation, whereas the frequency of La Niña events is associated with the Pacific decadal oscillation (Wang et al. 2009). In contrast, some studies have shown different types of El Niño (e.g., Fu et al. 1986, among many others), especially the conventional, canonical, or eastern Pacific type of El Niño (e.g., Kao and Yu 2009; Yeh et al. 2009) and the central Pacific type of El Niño (e.g., Kao and Yu 2009; Yeh et al. 2009; Lee and McPhaden 2010), date line El Niño (e.g., Larkin and Harrison 2005), El Niño Modoki (e.g., Ashok et al. 2007), or warm pool El Niño (Kug et al. 2009). It has been argued that the 1990–95 El Niño event can be alternatively classified as the nonconventional El Niño (e.g., Ashok et al. 2007; Kao and Yu 2009; Kug et al. 2009; Yeh et al. 2009; Lee and McPhaden 2010). It has also been argued that the nonconventional El Niño became more common during the late twentieth century, and that there will be an increased frequency of this type under a future anthropogenic global warming scenario (Yeh et al. 2009).
While changes in the properties of ENSO in connection with the global warming trend and lower-frequency variability have already been examined, much less attention has been paid to the effect of higher-frequency variability on ENSO frequency, especially the change of the annual cycle. Often, an El Niño–La Niña event is defined using traditional SST anomaly (SSTA) (e.g., Trenberth 1997), which is the departure from a mean annual cycle. This mean annual cycle is conventionally defined as the climatological monthly mean of the data for a given period, for example, 30 consecutive years, as suggested by the World Meteorological Organization (WMO). Such a conventional approach assumes that the annual cycle is perfectly repetitive from year to year. However, as reported by several recent studies (e.g., Gu and Philander 1995; Kim and Chung 2001; Pezzulli et al. 2005; Wu et al. 2008; Stine et al. 2009; Qian et al. 2011) that applied various time series analysis tools, the annual cycle of the tropical eastern Pacific SST and that of the surface air temperature (SAT) have large year-to-year differences. For example, Gu and Philander (1995) reported that the annual cycle (obtained using a 0.7–1.2-yr bandpassed wavelet filter) of the SST at 125° and 103°W along the equator is weak in El Niño years and strong in La Niña years. In addition to the year-to-year differences, the annual cycle of the tropical eastern Pacific SST has prominent interdecadal change (e.g., Gu and Philander 1995; Setoh et al. 1999; Torrence and Webster 1999). For example, Torrence and Webster (1999) reported that the annual cycle (1 yr) variance of the Niño-3 (5°S–5°N, 150°–90°W) SST is relatively larger during the period 1935–60. Changes in the annual cycle of the tropical Pacific SST are evident even if time series analysis tools are not used. For example, Xue et al. (2003) reported that the standard deviation of Niño-3 changes significantly from one 30-yr reference period to another. Moreover, differences in the mean annual cycles of SST for different 30-yr reference periods are also noticeable (e.g., Figs. 1 and 2), especially in the midlatitude Indian Ocean, the midlatitude North Pacific Ocean, the tropical eastern Pacific Ocean, and the mid–high-latitude North and South Atlantic Oceans (Fig. 1), and are not uniform throughout the year in the Niño-3 region (e.g., Fig. 2b). Thus, the conventional SSTA more or less retains a portion of the annual cycle, which may be judged an “El Niño–La Niña” event. For example, Pezzulli et al. (2005) suggested that the three cold ENSO episodes after 1998 were due to an increase in amplitude of seasonality rather than being three distinct La Niña events. Indeed, the different definitions of annual cycle, which serve as the reference frame for climate anomaly, could lead to completely different physical interpretations of climate phenomena and variability (Wu et al. 2008).
Differences between the mean annual cycle of 1971–2000 and that of 1961–90 for SST from (top left) January to (bottom right) December (ordered by row). Contour interval in all panels is 0.1°C.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
Different mean annual cycles of the Niño-3 SST relative to (a) different base periods and (b) with their mutual differences. (c) Different SSTAs for 1950–79 with respect to (w.r.t.) 1950–79 base period, 1961–90 base period, and 1971–2000 base period. Black dashed lines indicate ±0.5°C.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
Another complication related to the annual cycle is that updating the reference period from 1950–79 (e.g., Trenberth 1997) to 1961–90, and recently to 1971–2000, as suggested by the WMO and by Xue et al. (2003), results in shifts in the reference mean annual cycle in the tropical eastern Pacific SST, for example, Niño-3 SST (Fig. 2). These shifts occur not only in the mean value but also in different months with different magnitudes (Fig. 2b). Updating the reference periods therefore result in different SSTAs, which would potentially affect the classifications of El Niño–La Niña events as well as the estimations of their strength (e.g., Fig. 2c), although the underlying physical process of past ENSOs is already a reality. This inconsistency has made ENSO monitoring and forecasting, as well as research on ENSO property changes, more difficult.
This study aims to answer the following questions in the reference frame of the amplitude–frequency modulated annual cycle (MAC) (Wu et al. 2008) and in the framework of ENSO being the interannual variability of the eastern tropical Pacific SST: (i) What has led to changes in the frequency of ENSO events in the eastern tropical Pacific? Are the changes due to the changes in the annual cycle and/or in the background mean state, or to the change in the property of ENSO interannual variability itself? (ii) Why was there a prolonged 1990–95 El Niño (the anomalous 1990s) event? These questions are addressed through investigating the secular changes in the annual cycle, ENSO interannual variability, and the background mean state of the tropical eastern Pacific SST. Since several recent studies have argued that the circulation patterns and the mechanisms behind the two types of El Niño are quite different (e.g., Ashok et al. 2007; Weng et al. 2007; Kao and Yu 2009; Yu et al. 2010), this study applies more to the conventional El Niño, which is considered to be the warmer-than-normal sea surface temperature in the eastern tropical Pacific (Ashok and Yamagata 2009). Niño-3 is thus a desirable research index (e.g., Kao and Yu 2009; Yeh et al. 2009). In addition, the Niño-3 region has a well-defined annual cycle (e.g., Fu et al. 1986), which, compared to the Niño-3.4 or to the Niño-4 (5°S–5°N, 160°E–150°W) regions (figures not shown), explains a much more significant proportion of the total variance of SST. This study takes advantage of a recently developed adaptive and temporally local data analysis method—the ensemble empirical mode decomposition (EEMD) (Wu and Huang 2009; Huang and Wu 2008). The results revealed in this study call for attention to the potential impact of changes in the annual cycle.
The remaining sections of the paper are organized as follows: Section 2 describes data and analysis methods, and includes the classification criteria of the ENSO warm–cold episodes. Results are described in section 3, where the validation of our methods is demonstrated. Discussions about the classification of El Niño–La Niña events and possible reasons for changes in the amplitude of ENSO interannual variability are presented in section 4, followed by a summary and conclusions in section 5.
2. Data and methods
The monthly National Oceanic and Atmospheric Administration (NOAA) extended reconstructed SST dataset version 3 (NOAA ERSST.V3) for the period 1854–2008 (Smith et al. 2008) is used. Because of the relatively poor quality of the dataset prior to 1879 (e.g., An and Choi 2009), we select the data from January 1880 to December 2008 for this study. The Niño-3 index, which is traditionally used in ENSO monitoring and forecasting, is obtained through averaging the SST over the Niño-3 region. The monthly sea level pressure (SLP) and geopotential height at 500 hPa (GHT500) from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996; Kistler et al. 2001) from 1951 to 2007 are also used.
The method used to adaptively and temporally locally decompose the Niño-3 SST index series is the recently developed data-adaptive filter for nonlinear and nonstationary time series analysis—the EEMD (Wu and Huang 2009; Huang and Wu 2008), which is the most recent improvement of the EMD method (Huang et al. 1998; Huang and Wu 2008). The effectiveness of the EMD–EEMD method has been recently documented in many geophysical applications (e.g., Huang and Wu 2008; Wu et al. 2008; Wu and Huang 2009; Love et al. 2008; Qian et al. 2009, 2010, 2011; Ruzmaikin and Feynman 2009; Franzke 2009, 2010; Franzke and Woolings 2011; Breaker and Ruzmaikin 2010). In particular, the capability of EEMD and the advantage of using it to extract the annual cycle component (which has strong amplitude–frequency modulation) from a climate variable have been validated through analyzing synthetic data, monthly SST data, and daily surface air temperature records (Wu et al. 2008; Qian et al. 2011). The EEMD is used in the present study to decompose the Niño-3 SST (not the anomaly) index series; the added white noise in each EEMD ensemble member has a standard deviation of 0.4 and an ensemble size of 1000 is used. The first component (C1) of direct output of the decomposition is taken as the “updated” C1 (C1 in Fig. 3a). To obtain a simple wave form of the annual cycle, a refining process that is described in Wu et al. (2008) is applied: the second and third components of the direct output are combined and then subjected to an additional EMD sifting. The first intrinsic mode function (IMF) (Huang and Wu 2008) resulting from this additional sifting is taken as the updated C2 and the MAC (Fig. 3); and the leftover is taken as the updated C3. The C4–C9 are the fourth to ninth components of EEMD direct output. The last component is the residual (“R” in Fig. 3a) and also the nonlinear trend. Based on the mean period (see section 3a) of each component, we then reconstruct the decomposition in Fig. 3a into five major time-scale components: taking C1 as the high-frequency component (HF in Fig. 3b, representing intra-annual variability); taking C2 as the MAC component (MAC in Fig. 3b); combining C3–C5 as the ENSO interannual time-scale component (ENSO in Fig. 3b, representing variation within a period of 2–7 yr); combining C6–C9 as the decadal component (“decadal” in Fig. 3b); and taking R as the trend (“trend” in Fig. 3b). To eliminate the minor influence of end effect on our quantified results, the first and last years of all the decomposed results are excluded, leaving the period 1881–2007 to be analyzed.
Decomposition and reconstruction of the Niño-3 SST index from January 1881 to December 2007. (a) Reprocessed EEMD components of the raw Niño-3 SST index. (b) Reconstruction results from components in (a), where “input” stands for the raw monthly Niño-3 SST index. In the ENSO panel, dashed lines indicate ±0.5°C. In the MAC and ENSO panels, blue, red, and green lines indicate input data from 1880 to 2008, from 1880 to 1998, and from 1880 to 1988, respectively. Plotting sequences in these two panels are blue, red, and green lines. Because of the temporally local characteristics of the EEMD method, the previous lines are almost covered by later plotting for the overlapped time spans.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
To validate the isolation of the annual cycle by the EEMD method, the Morlet wavelet analysis (Torrence and Compo 1998) is used to analyze the anomaly with respect to the MAC (MA), which is also compared with the traditional anomaly relative to the 1971–2000 reference period (TA) (Fig. 4). Correlation maps between the ENSO interannual component displayed in Fig. 3b and the global SST, SLP, and GHT500 are made to further validate the isolation of the ENSO interannual signal (Fig. 5). The classification of ENSO warm–cold episodes by the Climate Prediction Center (CPC) is based on a minimum of five consecutive overlapping seasons that exceed a threshold of ±0.5°C for the oceanic Niño index, a 3-month running mean of ERSST.v3 SSTA in the Niño-3.4 region relative to the 1971–2000 base period (details can be found at www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml). In this study, ENSO warm–cold episodes are alternatively classified as a minimum of five consecutive months that exceed a threshold of ±0.5°C for only the ENSO interannual component in Fig. 3b. On the basis of this classification (Table 1), the composites for the MACs for El Niño years, La Niña years, and the climatological mean are obtained (Fig. 6b). Variability in the amplitude of the MAC and in that of the ENSO component are obtained through a cubic spline fitting to the local maxima of 
A wavelet analysis (a) for Niño-3 SST anomaly w.r.t. the MAC, that is, departure from the MAC and (d) for the traditional SSTA, that is, departure from the 1971–2000 mean. (b),(e) Corresponding wavelet power spectra, and (c),(f) corresponding global wavelet spectra, where dashed lines indicate a significance level at p < 0.05.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
Correlations between the ENSO component in Fig. 3b and the monthly global (a) SST, (b) SLP, and GHT500 (c) fields from 1951 to 2007. Dark (light) shading denotes positive (negative) values, and the 0 contour is removed. Contour interval is 0.1.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
Classification of ENSO events from 1881 to 2007. Weak (W): peak strength reaches 0.5°–1°C; moderate (M): 1°–1.5°C; and strong (S): >1.5°C. Asterisk means analyzed only from Jan 1881 to Dec 2007.
(a) MAC of the Niño-3 SST for every year from 1881 to 2007, together with the mean MAC of the entire series. (b) Composites of the MAC of the Niño-3 SST for the period 1881–2007 during warm (El Niño) and cold (La Niña) event years according to the classification in this study, where solid, circle, and dotted lines indicate the mean MAC, the composite of the MAC in warm years, and the composite of the MAC in cold years, respectively. (c),(d) Amplitude change of the MAC and of the ENSO component, respectively, together with their secular trends (dashed lines) fitted by EEMD. (e) Phase (timing of the annual cycle transition from the cold to the warm phase) change of the MAC, together with its linear trends (dashed lines) for the periods 1881–1938 and 1939–2007.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
3. Results
a. Validation of the MAC and ENSO interannual components
The mean periods for the EEMD components are 4.2 months for C1, 12 months for C2, 2.0 yr for C3, 3.6 yr for C4, 6.0 yr for C5, and 13.6 yr for C6. The quasi-decadal component (C6 in Fig. 3a) is consistent with the findings of Weng (2005), who reported that the cycle length of the decadal signal of the Niño-3.4 SSTA varies between 10 and 14 yr, and is also consistent with the findings of Wu and Huang (2009), who identified a 10.5-yr component from the Southern Oscillation index (SOI). The reconstruction results (Fig. 3b) show that the total variance of the Niño-3 SST index is explained 56.8% by the MAC, 28.7% by ENSO, and 3.0% by decadal, indicating that the MAC and ENSO are the dominant components of the SST in the Niño-3 region. To illustrate the “stability” and data-length independence of the MAC and ENSO reconstruction, we apply the decomposition method to two shortened versions of the same Niño-3 SST, one ending in 1997 and the other ending in 1987, to mimic the scenario in which we analyze the same SST one or two decades earlier. The results of the different decomposition and reconstructions for the MAC and ENSO are also shown in Fig. 3b. The MAC and ENSO components resulting from these two shortened versions remain almost the same for the overlapped temporal span (with a negligibly small difference: the square root errors for the 1881–1997 interval and the 1881–1987 interval are only 0.016 and 0.019, respectively, for the MAC component; and 0.0265 and 0.0526, respectively, for the ENSO component, mostly from the minor end effect at the right end). These results confirm that the reconstructions of the MAC and ENSO for a given temporal episode can be “uniquely” defined, which bypasses the dilemma caused by the multiple versions of climate anomaly (and therefore multiple definitions of ENSO strength) for a given episode, for example, the case in Fig. 2c, following the WMO suggestion that the reference climatology is updated every decade as new decades of data are collected and integrated into the old data. Thus, the method in the present study facilitates attribution in the subsequence. The decadal component is in its negative phase from the early 1940s to the late 1970s and shifts to a positive phase in the late 1970s, which is consistent with the findings of many previous studies (e.g., Zhang et al. 1997) (Fig. 3b). The trend shows generally a monotonic warming trend during the entire period of 1881–2007, suggesting that the background climate state of the Niño-3 SST is warming.
To illustrate how the MAC obtained by using EEMD isolates temporally local variability of the quasi-annual cycle, the Morlet wavelet analysis is applied to the MA series (Fig. 4a). It is evident that MA contains almost no variability of a quasi-1-yr period throughout the data span (Fig. 4b), validating that the MAC has isolated the annual cycle of the original data cleanly. In contrast, when the same Morlet wavelet analysis is applied to TA (Fig. 4d), the result shows that TA still contains annual time-scale variability at many temporal locations throughout the data span (Fig. 4e). This residual annual cycle could potentially be judged as an El Niño–La Niña event, as will be discussed in section 3c, although it is not significant compared to the ENSO interannual time scale (Fig. 4f).
The ENSO component obtained in this study shares many characteristics with the conventional ENSO. The correlation maps between the ENSO component and the monthly global SST and SLP fields displayed in Fig. 5 show that, corresponding to the positive (negative) ENSO phase, there is a positive (negative) SST anomaly in the equatorial Indian Ocean and equatorial eastern Pacific Ocean SST; and that there is a negative SST (positive) anomaly in the Maritime Continent (Fig. 5a), together with a seesaw pattern (which resembles the Southern Oscillation) over the Maritime Continent and the eastern Pacific for SLP (Fig. 5b). The correlation map of ENSO and GHT500 shows the dominant feature of the zonally symmetric geopotential height variability in the tropics as well as wave trains originating in the central tropical Pacific propagating to high latitudes (Fig. 5c). These zonally symmetric GHT500 and wave trains have been previously well diagnosed (e.g., Yulaeva and Wallace 1994) and theoretically explained (e.g., Wu et al. 2001) as typical circulation patterns driven by the zonally asymmetrical thermal forcing associated with ENSO variability.
b. Characteristics of the MAC and ENSO interannual components
The climatological mean of the MAC of the Niño-3 region SST has a maximum in April and a minimum in October, with a 2.5°C maximum-to-minimum range (Fig. 6a). This characteristic agrees with the findings of Gu and Philander (1995), who analyzed the annual cycle of local SST at 125° and 103°W along the equator using a bandpassed (0.7–1.2 yr) wavelet analysis method. It is also consistent with the climatological mean annual cycle for the period 1881–2007, which reaches a maximum of 27.2°C in April and a minimum of 24.6°C in September/October (Fig. 2a). In April, the southeast trades in the eastern part of the tropical Pacific basin are weakest (Dijkstra 2006). The relaxation of the meridional wind and the large heat flux across the sea surface at that time contributes to the seasonal maximum in SST. During September and October, the strong upwelling due to strong southeast trades, together with a weak heat flux, results in a seasonal minimum in SST (Koeberle and Philander 1994; Gu and Philander 1995).
The MACs of individual years have large year-to-year variation (Fig. 6a). For each month, the year-to-year difference can range from 1.2° (April) to 2.0°C (November). For example, the MAC can be 0.7°C in April but reach −1.9°C in October. Composites of MACs during El Niño and La Niña years (Fig. 6b) show that the amplitude of the MAC tends to be smaller than the climatological mean MAC during El Niño years and larger during La Niña years. This result confirms the results of Gu and Philander (1995) on the interaction between the SST annual cycle and ENSO. From the composite MACs, we also find that while the peaks of the mean MAC for El Niño years and for La Niña years are both in April, the troughs of the mean MAC for El Niño years appear one month earlier than in the climatological mean MAC, whereas the troughs of the mean MAC for La Niña years appear one month later than in the climatological mean MAC, which has not been noted before. This difference in the timing of the minima between an El Niño year and a La Niña year might be attributed to the weakening upwelling commencing earlier in El Niño years than in La Niña years. A more detailed mechanism remains to be worked out.
The large amplitude modulation displayed in Fig. 6a leads to further investigation of the variability and change in the instantaneous amplitude of the MAC for the period 1881–2007 (Fig. 6c). The amplitude of the MAC exhibits large interannual variability during this period: the mean amplitude is 1.3°C, and the largest amplitude is 1.95°C in 1956 and the smallest is 0.77°C in 2003. The amplitude in 1997 is 0.82°C, the third smallest amplitude during this period. The overall linear trend for the period 1881–2007 is not statistically significant (0.007°C century−1); however, the adaptive centennial nonlinear trend (Wu et al. 2007), which is the sum of the last three components extracted using the EEMD method, reveals a clear secular change from an increasing amplitude to a decreasing one (Fig. 6c), with the turning point in the late 1940s (around 1947/48). This turning point is determined by identifying the maximum amplitude location (with its first-order derivative being zero) in the centennial trend of amplitude (figure not shown). The peak of this centennial trend is about 1.4°C, and the minimum is 1.2°C in the early 2000s, with a difference of 0.2°C and a decrease of 14% since the late 1940s. These characteristics are consistent with those identified by Torrence and Webster (1999), who analyzed the cumulative variance curve for the 1-yr Niño-3 SST variance using the wavelet method and reported higher annual cycle variance from 1935 to 1960. However, the secular change in the amplitude of the annual cycle of the Niño-3 SST is a new finding in this study: the linear trend for the period 1881–1948 is 0.028° and −0.024°C decade−1 for the period 1948–2007; both are statistically significant at p < 0.01 under the Mann–Kendall test.
The amplitude of the ENSO component also has strong interannual variability: the largest amplitude is 2.7°C in 1997 and the smallest is 0.025°C in 2000 (Fig. 6d). The adaptive centennial nonlinear trend (Wu et al. 2007) extracted using the EEMD method shows a clear secular change from a decreasing to an increasing amplitude with the turning point in the late 1930s (around 1937). This result is consistent with the findings of Torrence and Webster (1999), who analyzed the time series of 2–7-yr variance in the Niño-3 SST index using the wavelet method and reported intervals of high ENSO variance (1875–1920 and 1960–90) and an interval of low variance (1920–60). However, the finding that the amplitude decreasing trend (−0.037°C decade−1, p < 0.01) during the period 1881–1937 and the amplitude increasing trend (0.039°C decade−1, p < 0.05) during the period 1937–97 are both significant under the Mann–Kendall test is new to this study. Note that the trend for the period 1937–2007 is not significant. The EEMD centennial trend of amplitude of ENSO has amplified about 30% from the late 1930s to the present (from 0.7° in 1938 to 0.9°C in 2007), which is in line with the estimation by Zhang et al. (2008), who found a 60% amplifying in the last 50 yr by applying a bandpass filter of 5–85 months to the linearly detrended Niño-3 SSTA index for the years 1870–2006, although the original SST datasets used in their studies were different than ours. Further analysis in this study shows that the variability in the amplitude of the MAC and in that of ENSO are negatively correlated at r = −0.26 for the period 1881–2007. This negative correlation agrees with the results of Gu and Philander (1995) and Torrence and Webster (1999). Note that the turning point of secular change in the amplitude of ENSO is about 8–10 yr earlier than that of the MAC. The causes for these secular changes remain to be understood.
The secular change in the phase of the MAC is displayed in Fig. 6e. The transition from a cold phase to a warm one for the MAC usually happens around January, but it can be as early as the previous November and as late as the following February. The linear trend in the phase transition of the MAC shows a trend of delay for the period 1881–1938 at about 3.0 day decade−1, which is significant at p < 0.1 under the Mann–Kendall test, and no noticeable trend for the period 1939–2007. This transition point (in 1938) is determined as the temporal location of the peak of the Mann–Kendall statistics (figure not shown). Several previous studies have reported that for most of the temperate zones, the seasons occur later over the oceans (e.g., Stine et al. 2009) and earlier over land (e.g., Stine et al. 2009; Qian et al. 2011) in the past five decades or so. The differences in the phase change of the annual cycle over the tropical and the temperate oceans and those over the oceans and lands remain to be understood. Nevertheless, the phase of the SST annual cycle is not stable over many oceans.
It should be noted that the SST data are more reliable after 1950 than before 1950 because of the increased data coverage over the global ocean. Although we cannot determine whether the transition in the 1940s, from an increasing to a decreasing trend in the amplitude of annual cycle, is real or due to a change in SST data coverage based on in situ SST data, the change (decreasing trend) in the amplitude of the annual cycle is also evident when only 1950–2008 data are investigated (Fig. 7). The linear trend in the amplitude of the annual cycle is −0.016°C decade−1 (−0.017°C decade−1) for the period 1951–2007, the result of decomposing input data for the period 1880–2008 (1950–2008). Both trends are significant at p < 0.05 under the Mann–Kendall test. The reason for this consistency is that the EEMD method is a temporally local decomposition method; the annual cycle and its amplitude for the overlapped time span remain almost the same as when they are obtained from input data that are shortened or extended. These characteristics are also shown in Fig. 3b. These experiments add confidence to our result, especially after 1950, when the SST data are more reliable.
Test for data-length independence of the amplitude of the MAC. Solid blue line as in Fig. 6c, indicating the amplitude (solid lines) of the annual cycle of the Niño-3 SST index for the period 1881–2007. Dashed blue line indicates the linear trend for the period 1951–2007. Red lines indicate the input data from 1950 to 2008, whose results are plotted later than those of the input data from 1880 to 2008. Because of the temporally local property of EEMD method, the lines plotted earlier are almost covered.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
c. ENSO frequency and reasons for the prolonged 1990–95 El Niño event
With the effects of the annual cycle and the decadal or longer time-scale changes of SST removed, the warm (El Niño) or cold (La Niña) episodes of ENSO classified in this study represent only the interannual variability of SST. As shown in Table 1, there have been 30 El Niño events and 29 La Niña events from 1881 to 2007. There have been 5 strong El Niño years (1888, 1972, 1982, 1987, and 1997) and no strong La Niña years. The asymmetry of El Niño and La Niña events (the non-Gaussian or nonzero skewness characteristics of ENSO) revealed here is consistent with the findings of many previous studies (e.g., An and Choi 2009; Kao and Yu 2009), although the results have been obtained from different analysis methods. These characteristics, typical of an eastern Pacific type of El Niño (Kao and Yu 2009), may be attributed to nonlinear dynamical processes, which cause the positively skewed ENSO events by intensifying El Niño events and suppressing La Niña events (An and Choi 2009). Such nonlinear dynamical processes include anomalous thermal advections by anomalous three-dimensional currents (Jin et al. 2003; An and Jin 2004) and the similar asymmetric damping mechanism of ENSO by the tropical instability waves (An 2008). Since 1976, there have been seven El Niño events and six (including the one developing in 2007) La Niña events, suggesting that the frequency of El Niño events is almost equivalent to that of La Niña events, and ENSO still maintains its own oscillation at quasi 2–7 yr, although its amplitude has increased since the 1940s (Fig. 6d). The tendency for more El Niño and fewer La Niña events reported by previous studies (e.g., Trenberth and Hoar 1996) when El Niño and La Niña are defined on the basis of the reference line of time-averaged temperature (i.e., SSTA) appears to be a result of secular warming superimposed on a decadal transition from a negative phase to a positive one (Figs. 3b and 8a), which may not result solely from tropical air–sea interactive dynamics (e.g., Collins et al. 2010). Note that from 1990 to 1995, the El Niño event, as classified in this study, lasts for only one year (from July 1991 to June 1992; Table 1), suggesting that the prolonged 1990–95 El Niño event reported in previous studies (e.g., Trenberth and Hoar 1996) did not develop by ENSO interannual variability only. Most of the El Niño and La Niña events classified in this study are consistent with the CPC’s classification (the comparison interval is 1950–2007), thus providing a base for subsequent analyses to identify the effects of changes in annual cycle and decadal background warming on the prolonged 1990–95 El Niño event.
Comparisons between the ENSO component (thick red line) in Fig. 3b and the traditional Niño-3 SSTA relative to the 1971–2000 base period (thin blue line). Dotted lines indicate ±0.5°C. (a) Solid black line indicates the mean state, which is the combination of the decadal and trend components in Fig. 3b. (b) Part of (a), but for the 3-month running mean of TA, together with the combinations of different components, where ENSO, “D” (decadal), and “T” (trend) are as in Fig. 3b, and residual annual cycle (“resAC”) represents the difference between the MAC and the mean annual cycle for the period 1971–2000.
Citation: Journal of Climate 24, 24; 10.1175/JCLI-D-10-05012.1
Previous studies suggested that the prolonged persistence of warm conditions in the early 1990s (1990–95) was highly unusual (Trenberth and Hoar 1996) and as unexpected as the exceptional intensity of the El Niño events in 1982 and 1997 (Philander and Fedorov 2003). If the CPC definition for an El Niño–La Niña event is applied to the Niño-3 SST, then three El Niño events would be identified from 1990 to 1995 (Fig. 8b), that is, from May 1991 to June 1992, from March 1993 to July 1993, and from October 1994 to February 1995, which are close to the estimations by Trenberth and Hoar (1997) for Niño-3.4 SST. The answer to whether the three El Niño events should be regarded as one long El Niño or three events in succession (Trenberth and Hoar 1997) remains ambiguous. Here, we refer to it as a warm phase of ENSO, following the precedent of Trenberth and Hoar (1997). Clearly, this warm phase of ENSO is not reflected by the ENSO interannual variation only (Fig. 8b). Three synthetic experiments based on the sum of different filtered components of time series analysis are performed. Experiment 1 (the combination of ENSO and decadal components) shows that the periods exceeding 0.5°C for five consecutive months are from May 1991 to September 1992 and from January 1993 to September 1993, suggesting that the traditionally defined 1991/92 El Niño is attributable to the interannual variation and reinforced by a positive phase of decadal variation (mainly a quasi-decadal variation for this data span; see C6 in Fig. 3a), and that the traditionally defined 1993 El Niño is attributable to this positive phase of decadal variation. This result is somewhat consistent with the results of Latif et al. (1997) and Hasegawa and Hanawa (2006), in that there appears to be a decadal mode in the anomalous 1990s. The physical mechanism underlying quasi-decadal variation is not fully understood at present, but it has been identified in the tropical Pacific SST by many previous studies (e.g., Tourre et al. 2001; Hasegawa and Hanawa 2006). Experiment 2 (the combination of ENSO, decadal, and trend) shows that the period exceeding 0.5°C for five consecutive months is from May 1991 to October 1993, suggesting that the background secular warming superimposes on decadal variation to enhance and lengthen the 1991–93 El Niño. If the combination of decadal variation and secular warming is viewed as the background mean state, then the 1991/92 El Niño is attributable to an interannual variation superimposed by change in this mean state, and the 1993 El Niño is due to change in this mean state. Experiment 3 (the combination of ENSO, decadal, trend, and the residual annual cycle) shows that the 3-month running mean of TA is very close to the combination of ENSO interannual variation, decadal variation, secular warming, and the residual annual cycle (the difference between MAC and the mean annual cycle of the period 1971–2000), suggesting that a 3-month running mean of TA still contains a portion of the annual cycle. This experiment implies that a significant part of the 1994/95 El Niño is attributable to the residual annual cycle, which cannot be fully excluded by a 30-yr mean annual cycle approach (e.g., Fig. 4e), since the annual cycle has long-term modulation (Fig. 6c). This residual annual cycle also reinforced the 1991/92 and 1993 El Niño events.
4. Discussion
Trenberth and Hoar (1997), but not the CPC, identified the El Niño event in 1993, although they both applied the Niño-3.4 SSTA. Such a discrepancy is simply because the CPC chose 1971–2000 as the reference period and Trenberth and Hoar (1997) chose 1950–79. This discrepancy suggests the limitation of using a 30-yr base period as a reference for classification of El Niño–La Niña, since the dynamical process of the occurred El Niño was already a reality and thus did not change no matter which 30 yr were chosen as the base period. In addition, the interannual and secular modulations in the annual cycle also challenge the use of the mean annual cycle as the reference for studying the climate anomalies. As to the limitations of a conventional climate anomaly, some studies suggested that they can be alleviated to a large extent by applying the MAC to serve as reference frame for climate anomaly since the MAC is temporally local and not sensitive to the extension of data length (Wu et al. 2008; Qian et al. 2010), as also can be seen from the results of the experiments in Fig. 3b. Other filters that can isolate the annual cycle temporally locally, such as the X-11 approach used in Pezzulli et al. (2005), are also reasonable. The local property ensures that “the values obtained in any one year are not overly biased by events happening at other times that could be unrelated to events in that year” (Pezzulli et al. 2005, p. 85) or “later evolution cannot change the reality that happened earlier” (Wu et al. 2008, p. 827).
This study finds that the prolonged 1990–95 El Niño event reported in previous studies (e.g., Trenberth and Hoar 1996) is not caused by ENSO interannual variability only. This study also finds that the 1994/95 El Niño, as classified by the definition in the CPC, is attributable to the residual annual cycle. These two findings raise the question of how to define an El Niño–La Niña event. Although Trenberth (1997) has made significant contributions in defining El Niño, more discussion may be needed that considers secular changes in the annual cycle and mean state since the climate system is highly nonstationary. If an arbitrary choice of the time-averaged SST of the past is still viewed as a reference line or background state, which is constant, then there is the possibility of a permanent El Niño with continued global warming. Latif et al. (1997) have suggested that ENSO is certainly the strongest fluctuation on interannual time scales, but the decadal variability cannot be described exclusively as a decadal modulation of ENSO. Since the physical mechanisms for the annual cycle, ENSO interannual variability, and background mean state are probably different (e.g., Collins et al. 2010), they may need to be separated when classifying El Niño/La Niña events.
It has been shown in this study that both the amplitude of the MAC and that of ENSO interannual variability in the Niño-3 SST have significant secular changes and secular transitions, with the turning point of the latter about 8–10 yr earlier than that of the former. This result suggests a possibility that the secular transition in the latter may lead to the secular transition in the former, since there is strong interaction between the SST annual cycle and ENSO (e.g., Gu and Philander 1995; Jin et al. 1996; Torrence and Webster 1999). An and Choi (2009) suggested that the decadal changes in the seasonality of the asymmetry of ENSO may influence the decadal changes in the amplitude of the annual and semiannual cycles of the tropical eastern Pacific SST by a nonlinear process. What, then, has led to the secular transition in the amplitude of ENSO interannual variability and the amplitude increase since the late 1930s? According to studies that were based on analyzing Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) climate models, there could be an anthropogenic global warming signal (e.g., Zhang et al. 2008). However, the ENSO response to global warming differs strongly from model to model (e.g., Hu et al. 2001; van Oldenborgh et al. 2005; Guilyardi 2006; Latif and Keenlyside 2009), partly because different climate models reproduce different background mean states for the future (Zhang et al. 2008). In contrast, secular changes and the transition in the amplitude of ENSO interannual variability could also be internal. For example, Cobb et al. (2003) suggested from analyzing fossil-coral-reconstructed ENSO proxy data that there is a case when the most intense ENSO (2–7-yr bandpass) activity occurred with little change in mean climate conditions. A nonlinear mechanism has previously been proposed for explaining ENSO amplitude modulations (Timmermann and Jin 2002). Given the limited length of the data record, it remains unclear whether the changes in the amplitude of the annual cycle or the ENSO interannual component are internal variability (having low-frequency cycles) or whether they are externally forced.
By analyzing the SSTA, some previous studies argued that the 1990–95 El Niño event can be alternatively classified as the nonconventional El Niño (e.g., Ashok et al. 2007; Kao and Yu 2009; Kug et al. 2009; Yeh et al. 2009; Lee and McPhaden 2010), so are some of the El Niño events among 2000–05. In this study, we identified 1991/92, 2002, and 2004 as weak El Niños in the Niño-3 region on the basis of the definition of ENSO being the interannual (2–7 yr) variability with the annual cycle cleanly excluded and by temperature threshold. Further investigations on spatial–temporal patterns and mechanisms of recent changes in El Niño events remain to be examined to better understand the conventional and nonconventional El Niños.
5. Summary and conclusions
The present study investigates changes in the frequency of ENSO events, especially the prolonged 1990–95 El Niño (the anomalous 1990s) event, in the context of secular change in the annual cycle, ENSO interannual variability, and background mean state of the tropical eastern Pacific SST. A nonlinear and nonstationary data analysis tool, the EEMD method, is applied to isolate those components from the Niño-3 SST index for the period 1880–2008. It is shown that the total variance of the Niño-3 SST index is explained 56.8% by the MAC and 28.7% by the ENSO interannual component. It is revealed that the MAC has large year-to-year variation instead of being close to its climatological mean. It is weaker than the climatology during an El Niño year and stronger than the climatology during a La Niña year. The peaks of the MAC appear in April in both El Niño and La Niña years, but the minima during an El Niño year appear one month earlier than the climatology and the minima of a La Niña year appear one month later than the climatology. In addition, the MAC has significant secular modulation in both amplitude and phase: there is a clear secular transition from increasing to decreasing in the amplitude around 1947/48, with both linear trends before and after this turning point significant (p < 0.01); the amplitude decreased by 14% between the late 1940s and 2007, and there is a significant (p < 0.1) delaying trend in the phase for the period 1881–1938 at about 3.0 day decade−1 but hardly any thereafter. It is found that ENSO interannual variability also has prominent secular modulation: there is a clear secular transition from decreasing to increasing in the amplitude around 1937, about 8–10 yr earlier than the secular transition of the MAC, with significant linear trends for the period 1881–1937 (p < 0.01) and for the period 1937–97 (p < 0.05). However, although the amplitude of ENSO interannual variability has increased by about 30% since the late 1930s, the frequency of El Niño events has been almost equivalent to that of La Niña events since 1976, and ENSO still maintains its own oscillation at quasi 2–7 yr, when El Niño events are represented as the collective variability of interannual time scales. The tendency for more El Niño events and fewer La Niña events, when ENSO events are classified on the basis of the reference line of time-averaged temperature, appears to be a result of change in the background mean state when it is viewed as the secular warming superimposed on a decadal variation. By proposing a synthetic experiment method based on time series analysis, it is further found that the prolonged 1990–95 El Niño event SST anomalous warming reported in previous studies (e.g., Trenberth and Hoar 1996) was not caused by ENSO interannual variability alone. Instead, it was caused by an interannual variation superimposed by a change in the background mean state for the 1991/92 warm period; by a change in the mean state for the 1993 warm period; and by the residual annual cycle, which cannot be fully excluded by a 30-yr mean annual cycle approach, for the 1994/95 El Niño (according to the classification by the CPC). Through this approach, we can easily identify which time-scale variability leads to a given “El Niño/La Niña event” classified by CPC. This study also calls for attention to the potential impact of changes in the annual cycle.
Acknowledgments
This research was jointly sponsored by R&D Special Fund for Public Welfare Industry (meteorology) (Grant GYHY201006022), the National Natural Science Foundation of China (Grant 41005039), the “Strategic Priority Research Program” of the Chinese Academy of Sciences (Grant XDA05090103), and LED/SCSIO/CAS open research program (Grant LED1005). Wu was sponsored by the National Science Foundation of the United States (Grant ATM-0917743), and Wang was sponsored by the National Science Foundation of China (Grant 40625017). NOAA ERSST.V3 data are provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, United States, from their Web site (www.esrl.noaa.gov/psd/data/gridded/data.noaa.ersst.html). The authors thank Prof. John M. Wallace (University of Washington, United States) for his helpful comments. We thank Ms. Kathy Fearon of the Florida State University for helping polish the manuscript and three anonymous reviewers for their many constructive suggestions.
REFERENCES
An, S.-I., 2008: Interannual variations of the tropical ocean instability wave and ENSO. J. Climate, 21, 3680–3686.
An, S.-I., and F.-F. Jin, 2004: Nonlinearity and asymmetry of ENSO. J. Climate, 17, 2399–2412.
An, S.-I., and J. Choi, 2009: Seasonal locking of the ENSO asymmetry and its influence on the seasonal cycle of the tropical eastern Pacific sea surface temperature. Atmos. Res., 94, 3–9.
Ashok, K., and T. Yamagata, 2009: The El Niño with a difference. Nature, 461, 481–484.
Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata, 2007: El Niño Modoki and its possible teleconnection. J. Geophys. Res., 112, C11007, doi:10.1029/2006JC003798.
Breaker, L. C., and A. Ruzmaikin, 2010: The 154-year record of sea level at San Francisco: Extracting the long-term trend, recent changes, and other tidbits. Climate Dyn., 36, 545–559, doi:10.1007/s00382-010-0865-4.
Cobb, K. M., C. D. Charles, H. Cheng, and R. L. Edwards, 2003: El Niño/Southern Oscillation and tropical Pacific climate during the last millennium. Science, 424, 271–276.
Collins, M., and Coauthors, 2010: The impact of global warming on the tropical Pacific Ocean and El Niño. Nat. Geosci., 3, 391–397.
Dijkstra, H. A., 2006: The ENSO phenomenon: Theory and mechanisms. Adv. Geosci., 6, 3–15.
Fedorov, A. V., and S. G. Philander, 2000: Is El Niño changing? Science, 288, 1997–2002.
Fedorov, A. V., and S. G. Philander, 2001: A stability analysis of tropical ocean–atmosphere interactions: Bridging measurements and theory for El Niño. J. Climate, 14, 3086–3101.
Franzke, C., 2009: Multi-scale analysis of teleconnection indices: Climate noise and nonlinear trends. Nonlinear Processes Geophys., 16, 65–76.
Franzke, C., 2010: Long-range dependence and climate noise characteristics of Antarctic temperature data. J. Climate, 23, 6074–6081.
Franzke, C., and T. Woollings, 2011: On the persistence and predictability properties of North Atlantic climate variability. J. Climate, 24, 466–472.
Fu, C., H. Diaz, and J. Fletcher, 1986: Characteristics of the response of sea surface temperature in the central Pacific associated with warm episodes of the Southern Oscillation. Mon. Wea. Rev., 114, 1716–1739.
Gu, D., and S. G. H. Philander, 1995: Secular change of annual and interannual variability in the tropics during the past century. J. Climate, 8, 864–876.
Gu, D., and S. G. H. Philander, 1997: Interdecadal climate fluctuations that depend on exchanges between the tropics and extratropics. Science, 275, 805–807.
Guilyardi, E., 2006: El Niño–mean state–seasonal cycle interactions in a multi-model ensemble. Climate Dyn., 26, 329–348.
Harrison, D. E., and N. K. Larkin, 1997: Darwin sea level pressure, 1876–1996: Evidence for climate change? Geophys. Res. Lett., 24, 1779–1782.
Hasegawa, T., and K. Hanawa, 2006: Impact of quasi-decadal variability in the tropical Pacific on ENSO modulations. J. Oceanogr., 62, 227–234.
Hu, Z.-Z., L. Bengtsson, E. Roeckner, M. Christoph, A. Bacher, and J. M. Oberhuber, 2001: Impact of global warming on the interannual and interdecadal climate modes in a coupled GCM. Climate Dyn., 17, 361–374.
Huang, N. E., and Z. Wu, 2008: A review on Hilbert-Huang transform: Method and its applications to geophysical studies. Rev. Geophys., 46, RG2006, doi:10.1029/2007RG000228.
Huang, N. E., Z. Shen, S. R. Long, M. C. Wu, E. H. Shih, Q. Zheng, C. C. Tung, and H. H. Liu, 1998: The empirical mode decomposition and the Hilbert spectrum for nonlinear and nonstationary time series analysis. Proc. Roy. Soc. London, 454A, 903–995.
Jin, F.-F., J. D. Neelin, and M. Ghil, 1996: El Niño/Southern Oscillation and the annual cycle: Subharmonic frequency-locking and aperiodicity. Physica D, 98, 442–465.
Jin, F.-F., S.-I. An, A. Timmermann, and J. Zhao, 2003: Strong El Niño events and nonlinear dynamical heating. Geophys. Res. Lett., 30, 1120, doi:10.1029/2002GL016356.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kao, H.-Y., and J.-Y. Yu, 2009: Contrasting eastern-Pacific and central-Pacific types of ENSO. J. Climate, 22, 615–632.
Kendall, M. G., 1955: Rank Correlation Methods. 2nd ed. Charles Griffin and Company, 196 pp.
Kim, K.-Y., and C. Chung, 2001: On the evolution of the annual cycle in the tropical Pacific. J. Climate, 14, 991–994.
Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247–267.
Koeberle, C., and S. G. H. Philander, 1994: On the processes that control seasonal variations of sea surface temperatures in the tropical Pacific Ocean. Tellus, 46A, 481–496.
Kug, J.-S., F.-F. Jin, and S.-I. An, 2009: Two types of El Niño events: Cold tongue El Niño and warm pool El Niño. J. Climate, 22, 1499–1515.
Larkin, N. K., and D. E. Harrison, 2005: On the definition of El Niño and associated seasonal average U.S. weather anomalies. Geophys. Res. Lett., 32, L13705, doi:10.1029/2005GL022738.
Latif, M., and N. S. Keenlyside, 2009: El Niño/Southern Oscillation response to global warming. Proc. Natl. Acad. Sci. USA, 106, 20 578–20 583, doi:10.1073/pnas.0710860105.
Latif, M., R. Kleeman, and C. Eckert, 1997: Greenhouse warming, decadal variability, or El Niño? An attempt to understand the anomalous 1990s. J. Climate, 10, 2221–2239.
Lee, T., and M. J. McPhaden, 2010: Increasing intensity of El Niño in the central-equatorial Pacific. Geophys. Res. Lett., 37, L14603, doi:10.1029/2010GL044007.
Love, B. S., A. J. Matthews, and G. J. Janacek, 2008: Real-time extraction of the Madden–Julian oscillation using empirical mode decomposition and statistical forecasting with a VARMA model. J. Climate, 21, 5318–5335.
Mann, H. B., 1945: Nonparametric tests against trend. Econometrica, 13, 245–259.
Pezzulli, S., D. B. Stephenson, and A. Hannachi, 2005: The variability of seasonality. J. Climate, 18, 71–88.
Philander, S. G., and A. V. Fedorov, 2003: Is El Niño sporadic or cyclic? Annu. Rev. Earth Planet. Sci., 31, 579–594.
Qian, C., C. Fu, Z. Wu, and Z. Yan, 2009: On the secular change of spring onset at Stockholm. Geophys. Res. Lett., 36, L12706, doi:10.1029/2009GL038617.
Qian, C., Z. Wu, C. Fu, and T. Zhou, 2010: On multi-timescale variability of temperature in China in modulated annual cycle reference frame. Adv. Atmos. Sci., 27, 1169–1182, doi:10.1007/s00376-009-9121-4.
Qian, C., C. Fu, Z. Wu, and Z. Yan, 2011: The role of changes in the annual cycle in earlier onset of climatic spring in northern China. Adv. Atmos. Sci., 28, 284–296, doi:10.1007/s00376-010-9221-1.
Rajagopalan, B., U. Lall, and M. A. Cane, 1997: Anomalous ENSO occurrences: An alternative view. J. Climate, 10, 2351–2357.
Ruzmaikin, A., and J. Feynman, 2009: Search for climate trends in satellite data. Adv. Adapt. Data Anal., 1, 667–679.
Setoh, T., S. Imawaki, A. Ostrovskii, and S.-I. Umatani, 1999: Interdecadal variations of ENSO signals and annual cycles revealed by wavelet analysis. J. Oceanogr., 55, 385–394.
Smith, T. M., R. W. Reynolds, T. C. Peterson, and J. Lawrimore, 2008: Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006). J. Climate, 21, 2283–2296.
Stine, A. R., P. Huybers, and I. Y. Fung, 2009: Changes in the phase of the annual cycle of surface temperature. Nature, 457, 435–440.
Timmermann, A., and F.-F. Jin, 2002: A nonlinear mechanism for decadal El Niño amplitude changes. Geophys. Res. Lett., 29, 1003, doi:10.1029/2001GL013369.
Torrence, C., and G. P. Compo, 1998: A practical guide to wavelet analysis. Bull. Amer. Meteor. Soc., 79, 61–78.
Torrence, C., and P. J. Webster, 1999: Interdecadal changes in the ENSO–monsoon system. J. Climate, 12, 2679–2690.
Tourre, Y. M., B. Rajagopalan, Y. Kushnir, M. Barlow, and W. B. White, 2001: Patterns of coherent decadal and interdecadal climate signals in the Pacific basin during the 20th century. Geophys. Res. Lett., 28, 2069–2072.
Trenberth, K. E., 1997: The definition of El Niño. Bull. Amer. Meteor. Soc., 78, 2771–2777.
Trenberth, K. E., and T. J. Hoar, 1996: The 1990–1995 El Niño–Southern Oscillation event: Longest on record. Geophys. Res. Lett., 23, 57–60.
Trenberth, K. E., and T. J. Hoar, 1997: El Niño and climate change. Geophys. Res. Lett., 24, 3057–3060.
van Oldenborgh, G. J., S. Y. Philip, and M. Collins, 2005: El Niño in a changing climate: A multi-model study. Ocean Sci., 1, 81–95.
Wang, X., D. Wang, and W. Zhou, 2009: Decadal variability of twentieth-century El Niño and La Niña occurrence from observations and IPCC AR4 coupled models. Geophys. Res. Lett., 36, L11701, doi:10.1029/2009GL037929.
Weng, H., 2005: The influence of the 11 yr solar cycle on the interannual–centennial climate variability. J. Atmos. Sol.-Terr. Phys., 67, 793–805.
Weng, H., K. Ashok, S. K. Behera, S. A. Rao, and T. Yamagata, 2007: Impacts of recent El Niño Modoki on dry/wet conditions in the Pacific rim during boreal summer. Climate Dyn., 29, 113–129.
Wu, Z., and N. E. Huang, 2009: Ensemble empirical mode decomposition: A noise-assisted data analysis method. Adv. Adapt. Data Anal., 1, 1–41.
Wu, Z., E. S. Sarachik, and D. S. Battisti, 2001: Thermally driven tropical circulations under Rayleigh friction and Newtonian cooling: Analytic solutions. J. Atmos. Sci., 58, 724–741.
Wu, Z., N. E. Huang, S. R. Long, and C.-K. Peng, 2007: On the trend, detrending, and variability of nonlinear and nonstationary time series. Proc. Natl. Acad. Sci. USA, 104, 14 889–14 894.
Wu, Z., E. K. Schneider, B. P. Kirtman, E. S. Sarachik, N. E. Huang, and C. J. Tucker, 2008: The modulated annual cycle: An alternative reference frame for climate anomalies. Climate Dyn., 31, 823–841.
Xue, Y., T. M. Smith, and R. W. Reynolds, 2003: Interdecadal changes of 30-yr SST normals during 1871–2000. J. Climate, 16, 1601–1612.
Yeh, S.-W., J.-S. Kug, B. Dewitte, M.-H. Kwon, B. P. Kirtman, and F.-F. Jin, 2009: El Niño in a changing climate. Nature, 461, 511–514, doi:10.1038/nature08316.
Yu, J. Y., H.-Y. Kao, and T. Lee, 2010: Subtropics-related interannual sea surface temperature variability in the central equatorial Pacific. J. Climate, 23, 2869–2884.
Yulaeva, E., and J. M. Wallace, 1994: The signature of ENSO in global temperature fields derived from the Microwave Sounding Unit. J. Climate, 7, 1719–1736.
Zhang, Q., Y. Guan, and H. Yang, 2008: ENSO amplitude change in observation and coupled models. Adv. Atmos. Sci., 25, 331–336.
Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10, 1004–1020.