Abstract

This study introduces a new methodology for identifying El Niño and La Niña events. Sea surface temperature (SST) anomaly patterns for El Niño and La Niña onset, peak, and end phases are classified by self-organizing maps (SOM) analysis. Both onset and end phases for El Niño and La Niña exhibit eastern Pacific (EP) and central Pacific (CP) types. The SST anomaly patterns in peak phase can be classified into EP, EP-like, and CP types for El Niño, and EP, mixed (MIX), and CP types for La Niña.

The general type of each El Niño or La Niña event is then defined according to the SST type for each of the three phases. There is no robust connection between the general types of the contiguous El Niño and La Niña except that the MIX La Niña rarely induces a subsequent CP El Niño. However, there are strong relationships between the end-phase type of El Niño and the onset-phase type of the subsequent La Niña. The EP-end-type El Niño favors transition to the CP-onset-type La Niña, while the CP-end-type El Niño favors transition to the EP-onset-type La Niña. On the other hand, the CP-end-type La Niña favors transition to EP-onset-type El Niño. Furthermore, an El Niño that occurs after the decay of La Niña favors initiating as an EP-onset type. These relationships are driven by different atmosphere–ocean dynamics, such as coupled air–sea feedback, thermocline feedback, slow SST mode, and Bjerknes feedbacks.

1. Introduction

El Niño–Southern Oscillation (ENSO) is the dominant component of tropical interannual variability, and affects weather and climate on a global scale. As early as the late 1980s, scientists noticed two types of developing processes associated with El Niño events. In one type, the emergence of SST warming off the South American coast leads that in the central Pacific by a few months, while in the other type SST warming off the South American coast lags that in the central Pacific (Fu 1985). Many recent studies have used the distribution of SST anomalies to identify two types of El Niño (Ashok et al. 2007; Kao and Yu 2009; Kug et al. 2009; Yeh et al. 2009). The conventional El Niño, with maximum SST anomalies located in the equatorial eastern Pacific, is referred to as the cold tongue El Niño (Kug et al. 2009, 2010) or eastern Pacific (EP) El Niño (Kao and Yu 2009; Yeh et al. 2009; Yu and Kim 2010). The newly discovered El Niño with maximum warming in the central equatorial Pacific is known as the date line El Niño (Larkin and Harrison 2005), El Niño Modoki (Ashok et al. 2007; Weng et al. 2007), warm pool El Niño (Kug et al. 2009, 2010), or central Pacific (CP) El Niño (Kao and Yu 2009; Yeh et al. 2009; Yu and Kim 2010).

These two types of El Niño have differing frequencies of occurrence, subsurface structures, and physical processes (Kao and Yu 2009; Kug et al. 2009; Xiang et al. 2013). They also have different impacts on dry/wet conditions in the Pacific rim (Weng et al. 2007), the climate of East Asia (Yuan and Yang 2012), tropical cyclone genesis over the South China Sea (Wang et al. 2014), global temperature (Banholzer and Donner 2014), and variability of the Indian Ocean dipole (Wang and Wang 2014). However, there is still no consensus on the classification of an El Niño event. Some studies indicate that some events may have mixed features from both EP and CP El Niño types (Kug et al. 2009; Yu et al. 2011; Yu and Kim 2013), and other research has further classified CP El Niño into two types based on their different impacts (Wang and Wang 2013).

La Niña events have received less attention than El Niño, and it is not clear whether there is more than one type of La Niña. Some studies show no significant difference among La Niña events (Kug et al. 2009; Kug and Ham 2011; Ren and Jin 2011), whereas others argue for the existence of two types of La Niña (Ashok et al. 2007; Cai and Cowan 2009; Kao and Yu 2009; Shinoda et al. 2011; Yu et al. 2011) with differing impacts on the tropical atmosphere (Yuan and Yan 2013) and tropical cyclone activity (Wang et al. 2013), as well as the North Atlantic Oscillation during the boreal winter (Zhang et al. 2015). The variability of La Niña events therefore needs further exploration.

In addition, it should be noted that previous studies of ENSO diversity have mainly focused on different spatial patterns of SST anomaly during the peak phase. However, the onset and decay phases of ENSO also play a very important role in forcing climate variability, including the timing of the East Asian monsoon (Wu and Wang 2002), hurricanes (Larson et al. 2012), and terrestrial rainfall (Lee et al. 2014b). Thus, more attention should be given to understanding the onset and end phases of ENSO.

The phase transitions between El Niño and La Niña in the ENSO cycle have also received much attention in recent decades. In a traditional ENSO cycle, negative (positive) subsurface ocean temperature anomalies propagate eastward along the thermocline, emerging on the surface of the eastern Pacific and finally reversing the ENSO phase in the transition from El Niño (La Niña) to La Niña (El Niño) (Li and Mu 1999, 2002; Chao et al. 2002). Several hypotheses have been presented to explain this phase transition, such as the delayed-oscillator theory (Schopf and Suarez 1988; Battisti and Hirst 1989), a slow SST mode (Neelin 1991; Wang and Weisberg 1994), recharge–oscillator mechanisms (Jin 1997a,b), and a stationary SST mode (Li 1997). However, further investigation is required to understand whether the different El Niño types undergo the same phase transition. Some studies suggest that there is no identifiable phase-reversal signal in the CP-type ENSO because it does not behave like a cycle or oscillation (Kao and Yu 2009). Some others argue that the interannual variability of upper-ocean heat content still exhibits oscillation in the warm pool ENSO due to recharge–discharge processes (Ren and Jin 2013). However, the thermocline oscillation is much stronger in the evolution of the EP El Niño than that of the CP El Niño (Li and Li 2014).

If the phase transition does exist in different types of El Niño and La Niña, it would be meaningful to know whether a certain type of La Niña (El Niño) is likely to follow a certain type of El Niño (La Niña). Yu et al. (2011) found that strong CP La Niña events tend to occur after strong EP El Niño events, while strong EP El Niño events do not always follow strong CP La Niña events. Actually, some strong La Niña events are also followed by a weak or moderate El Niño. Lee et al. (2014a) provided new insight into inter–El Niño variability, revealing some important features of the transition from El Niño to La Niña. Their results suggest that early-onset El Niño and weak El Niño both favor a transition to La Niña, but they did not focus on the transition from La Niña to El Niño or the spatial characteristics of El Niño and La Niña. Therefore, further investigation is required to understand the relationships between the types of El Niño and that of their contiguous La Niña.

To classify El Niño events into different types, various pairs of indices have been defined in previous studies using SST (Kao and Yu 2009; Kim et al. 2009; Kug et al. 2009; Yeh et al. 2009; Ren and Jin 2011; Takahashi et al. 2011; Yuan and Yang 2012) and subsurface ocean temperature (Yu et al. 2011). These indices may reveal some of the distinct properties of different types of El Niño, but do not result in consistent classification. This is primarily due to the fact that the indices are to a certain extent subjective and not necessarily optimal for describing the various El Niño flavors (Johnson 2013). All indices are based on average regional SST or subsurface ocean temperature anomalies rather than their spatial distribution in the tropics. Furthermore, different studies define the mature phase of El Niño differently, such as from December to February, from September to November, or from September to February, resulting in inconsistencies in the identification of these events. Recently, studies have attempted to classify El Niño events based on the pattern correlation method (Yu and Kim 2013), the distribution of sea surface salinity (Singh et al. 2011), or their different impacts on rainfall and typhoon tracks in southern China (Wang and Wang 2013).

In this study, a new perspective and methodology are introduced to identify the spatial types of El Niño and La Niña based on the self-organizing maps (SOM) technique. The method is used to classify the events not only in their peak phase, but also in their onset and end phases. The main purpose of this paper is to provide new insights into ENSO classifications in different phases and explore the relationships between the types of contiguous El Niño and La Niña in ENSO cycles.

The data and methodology used in this analysis are described in section 2. Results are presented in section 3, including patterns of different El Niño and La Niña phases identified by SOM. The relationships between contiguous El Niño and La Niña are further investigated in this section. A summary and conclusions are given in section 4.

2. Data and methodology

a. Data

Two SST datasets are used in this study. The first is the gridded monthly Extended Reconstructed Sea Surface Temperature, version 3b (ERSSTv3b), available beginning from 1853 from the National Oceanic and Atmospheric Administration (NOAA) with a horizontal resolution of 2° × 2°. This dataset was constructed using the most recently available International Comprehensive Ocean–Atmosphere Dataset (ICOADS) SST data and improved statistical methods that allow stable reconstruction using sparse data (Smith et al. 2008). The other dataset used to validate the results of the study is the sea ice and sea surface temperature data from the Met Office Hadley Centre (HadISST) with a horizontal resolution of 1° × 1°. This dataset was reconstructed using a two-stage reduced-space optimal interpolation procedure, followed by superposition of quality-improved gridded observations onto the reconstructions to restore local detail (Rayner et al. 2003). The analysis period of the study is 1951–2011, as high-quality data from both sources are available for this time span. The Simple Ocean Data Assimilation (SODA) ocean reanalysis (Giese and Ray 2011) with a horizontal resolution of 0.5° × 0.5° is used to derive surface wind stress fields and sea surface height (SSH), which is highly correlated with thermocline depth.

b. Definition of El Niño and La Niña events

The term “El Niño” has been traditionally associated with an unusual SST warming in the cold-water region along the Peruvian coast, but is now more generally used to describe large-scale oceanic warming in the central and eastern equatorial Pacific. The term “La Niña” is associated with the opposite phenomenon, which consists of a basinwide ocean surface cooling in the tropical Pacific. Various methods have been proposed to define El Niño and La Niña events, but the diversity in the characteristics of these events makes this a difficult task. Various kinds of Niño indices, such as Niño-1 + 2, Niño-3, Niño-3.4, and Niño-4, have been developed with different criteria to identify events based on average SST over different regions (Aceituno 1992; Trenberth 1997; Larkin and Harrison 2005; Kim et al. 2009; Kug et al. 2009; Yeh et al. 2009; Hu et al. 2012).

El Niño events can be defined as periods when SST in the central or eastern Pacific is warmer than the long-term climatology. The climatological SST in the central or eastern Pacific is usually calculated by the mean of the three most recent complete decades, such as 1971–2000 and 1981–2010. If the underlying trend toward increasing SST due to global warming were very weak, differences in the climatology of any 30-yr period would be very small, and there would be negligible impact on the identification of historic El Niño and La Niña events. However, numerous studies indicate that the oceanic temperature has significantly increased in the last 100 years (Levitus et al. 2000, 2012; Hansen et al. 2006; Gleckler et al. 2012), which means that the SST climatology for detecting El Niño and La Niña events has also shifted (L’Heureux et al. 2013). The climatology of monthly mean SST in the Niño-3.4 region for different 30-yr periods is shown in Fig. 1. This indicates that the increasing SST trend in the last century cannot be neglected. This trend also exists in other Niño regions (not shown). Thus, using the 1981–2010 climatology to identify historical El Niño events may result in less accuracy in the detection of El Niño and La Niña events. For example, because the historical SST climatology is lower than that in the period 1981–2010, the apparent strength of El Niño events decreases sharply and some events may not be identified at all.

Fig. 1.

Climatology of monthly mean SST (°C) for different time periods in the Niño-3.4 region.

Fig. 1.

Climatology of monthly mean SST (°C) for different time periods in the Niño-3.4 region.

The NOAA/Climate Prediction Center has developed a new strategy to account for the SST trend. Anomalies are obtained by removing the climatology of centered 30-yr periods, which are updated every five years. For example, SST anomalies from 1951 to 1955 are calculated by removing the climatology from 1936 to 1965, while for 1956–1960 the climatology covers the period from 1941 to 1970. This method can effectively remove the warming trend and interdecadal variability, and thus better reflects interannual ENSO variability. In the same way, monthly mean SST anomalies in this study are calculated by removing the corresponding climatology.

It is well known that SST anomalies vary not only among Niño-3, Niño-3.4, and Niño-4 regions in different types of El Niño and La Niña, but also during the evolution of a single event. To take advantage of all three indices to identify ENSO events, a new index is composed using the maximum amplitude of the three indices after the application of a 3-month running mean. An El Niño (La Niña) event is detected if the index is above 0.5°C (below −0.5°C) for at least five consecutive months (Fig. 2). The onset and end months are the first and the last months of the event period in which the amplitude of the index is above 0.5°C. The peak month occurs when the amplitude of the index reaches a local maximum (Fig. 2). It should be noted that some events, such as the 1998–2001 La Niña, have more than one peak. In this case, the peak with the maximum amplitude is the primary peak, and the others are treated as secondary peaks (Fig. 2).

Fig. 2.

An illustration of the identification of the El Niño and La Niña events as well as the time of their onset, peak, and end, respectively.

Fig. 2.

An illustration of the identification of the El Niño and La Niña events as well as the time of their onset, peak, and end, respectively.

c. Methodology

1) Self-organizing maps technique

The SOM technique is a type of unsupervised artificial neural network-based cluster analysis mainly used for pattern recognition and classification (Kohonen et al. 1995; Kohonen 2001). It is similar to k-means clustering to some extent, and follows the principle of all cluster analyses: to bring out M (MT) dominant patterns from total T fields while the M patterns can catch most of the spatial features of the T fields. It can preserve topology of the structure; that is, similar clusters are grouped together (similar patterns together) to facilitate easy visualization.

The SOM method has been widely applied in oceanographic and meteorological investigations to identify, for example, synoptic-scale atmospheric circulation patterns (Cassano et al. 2006; Johnson et al. 2008; Johnson and Feldstein 2010), intraseasonal variability of monsoons (Chattopadhyay et al. 2008), the evolution of the Madden–Julian oscillation (Chattopadhyay et al. 2013), SST patterns (Iskandar 2010), ENSO characteristics (Leloup et al. 2008; Johnson 2013), and features of ocean currents (Liu and Weisberg 2005; Liu et al. 2007; Jin et al. 2010). The advantages of this method include (i) more accuracy than leading modes of empirical orthogonal function (EOF) patterns for extracting nonlinear and asymmetric features (Iskandar 2009); (ii) a more robust approach than principal component analysis for extracting predefined patterns of variability; (iii) more accuracy and flexibility than other cluster methods, such as the k-means method (Lin and Chen 2006) and Ward’s method (Solidoro et al. 2007), in determining cluster membership; and (iv) a more intuitive approach than other methods because of its continuity in the data space.

In this study, we use the SOM method to identify SST anomaly patterns in the onset, peak, and end phases of El Niño as well as La Niña. Take the onset phase of El Niño as an example: Given the total number of El Niño events T during the period 1951–2011, there should be T SST anomalies fields with 81 × 25 grid points covering the tropical Pacific region over 25°S–25°N, 120°E–80°W. The shape and size of the network is predefined by taking the number of nodes (SST anomaly patterns) M into account. Each node on the network is associated with a weight vector with dimensions equal to that of the input vector. The weight vectors are assigned either random or linear starting values.

The next step is training the network to capture dominant SST anomaly patterns. For each input vector from the SSTA fields, the SOM searches for the best-matching node (BMN) on the predefined network based on minimum Euclidian distance. The node with the minimum distance to the input vector is identified as the BMN. This node and its neighboring nodes are trained by updating the weights so that they become closer to the input vector. The training process is repeated until the network converges. The weight vectors for the network (i.e., the SST anomaly patterns) are determined once the iteration is finished, and all T SST anomalies fields are then classified into M nodes.

Some parameters must be defined before running the SOM procedure. The “rectangular” lattice, the “sheet” map shape used to define the shape of the network, “linear” initialized method used to initialize weight vectors in the network, the “Gaussian” neighborhood function, and “batch” training used to update the weight vector for the best-matching node and its neighbors. The initial and final neighborhood radii are set to 3 and 0.001, respectively. The number of nodes M, or the number of SST anomaly patterns, is determined by the objective method illustrated in section 2c(2).

2) False discovery rate

The false discovery rate (FDR) method is a significance approach introduced to meteorology by Wilks (2006) and then successfully applied to distinguish significant differences between any two patterns in the SOM (Johnson 2013). The FDR refers to the expected proportion of local null hypotheses that are rejected but are actually true. In this study, the local hypotheses are that SST anomalies are significantly different between any two spatial patterns.

Taking the onset phases of El Niño as an example: If the number of SSTA patterns is set to K, there should be K(K − 1)/2 possible pairs of patterns to test the hypotheses described above. To determine the optimum number M, the SOM analysis is performed with the number of nodes K from 2 to N (NT). For each K, the significance test at 95% confidence is performed for the K(K − 1)/2 pairs of patterns, and the total number of the indistinguishable pairs is obtained. When K is set to 2, the number of statistically indistinguishable pairs of patterns is zero, but this increases to 1 when the K is set to 3. Thus, the optimum number of El Niño onset SST anomaly patterns is two.

3. Results

a. El Niño and La Niña events during 1951–2011

A total of 19 El Niño events and 15 La Niña events were identified from 1951 to 2011 (Table 1). Both El Niño and La Niña events tend to occur in the boreal late spring or summer, but some events actually initiated as early as early in March or even in January or as late as in boreal autumn. Although most events peak in the boreal winter, there are eight El Niño events and seven La Niña events that peak in autumn. Moreover, five El Niño events and five La Niña events end in winter rather than in the following spring or summer. Thus, both the boreal winter season [December–February (DJF)] and autumn season [September–November (SON)], commonly used in previous studies, have some deficiencies in identifying ENSO events and their peaks. In addition, it should be noted that two prolonged El Niño events (1968–70 and 1987–89) have two peaks. Thus, the 19 El Niño events have a total of 21 peaks. Similarly, the 1995–97 La Niña has two peaks while the 1973–75 and 1998–2001 La Niña events both have three peaks. Accordingly, the 15 La Niña events have a total of 20 peaks.

Table 1.

El Niño and La Niña events during 1951–2011 and the time and types of their onset, peak (primary) and end phases as well as their general types.

El Niño and La Niña events during 1951–2011 and the time and types of their onset, peak (primary) and end phases as well as their general types.
El Niño and La Niña events during 1951–2011 and the time and types of their onset, peak (primary) and end phases as well as their general types.

b. El Niño phase types

Figure 3 shows the number of indistinguishable pairs of El Niño and La Niña SST anomaly patterns at 95% significance level as a function of the number of SOM patterns during their onset, peak, and end phases, respectively. It indicates that the optimum numbers of patterns for both El Niño and La Niña in their onset, peak, and end phases are two, three, and two, respectively.

Fig. 3.

The number of indistinguishable pairs of SST anomalies patterns of the El Niño (solid line) and La Niña (dashed line) at the 95% significance level as a function of the number of SOM patterns during their (a) onset, (b) peak, and (c) end phases.

Fig. 3.

The number of indistinguishable pairs of SST anomalies patterns of the El Niño (solid line) and La Niña (dashed line) at the 95% significance level as a function of the number of SOM patterns during their (a) onset, (b) peak, and (c) end phases.

The two SST anomaly patterns for the onset phase of El Niño are shown in Figs. 4a and 4b. The anomalous warm SST in the first pattern appears in the eastern Pacific, especially off the Peruvian coast (Fig. 4a), while it emerges from the central Pacific in the second pattern (Fig. 4b). According to the location of maximum SST anomalies, the first pattern is thus referred to as the EP-onset type, while the second pattern is named as the CP-onset type. The corresponding BMN series (first row in Fig. 5) shows that the EP-onset type occurs more frequently than the CP-onset type before 1978, while the CP-onset type dominates after 1978. This result indicates that El Niño events were likely to originate in the eastern Pacific before 1978 but have tended to emerge from the central Pacific in more recent decades.

Fig. 4.

The horizontal patterns of monthly mean SST anomalies (color; °C) for the (top) onset, (middle) peak, and (bottom) end phases of El Niño. The frequency of occurrence for each type is labeled on the top of each panel.

Fig. 4.

The horizontal patterns of monthly mean SST anomalies (color; °C) for the (top) onset, (middle) peak, and (bottom) end phases of El Niño. The frequency of occurrence for each type is labeled on the top of each panel.

Fig. 5.

The classification for patterns of SST anomalies for each phase of El Niño (top three rows) and La Niña (bottom three rows) events. The three phases of the same event are linked by dotted line. The primary peak is marked by circle for the events with multiple peak phases.

Fig. 5.

The classification for patterns of SST anomalies for each phase of El Niño (top three rows) and La Niña (bottom three rows) events. The three phases of the same event are linked by dotted line. The primary peak is marked by circle for the events with multiple peak phases.

The three SST anomaly patterns for the peak phase of El Niño are shown in Figs. 4c–e. The first pattern, or EP-peak type, has the strongest SST anomalies, covering a wide region of the central and eastern Pacific, with a maximum off the Peruvian coast (Fig. 4c). The second pattern, or EP-like-peak type (Fig. 4d), exhibits positive SST anomalies in most of the central and eastern Pacific. This is similar to the first pattern, but SST anomalies are shifted westward to the central Niño-3 region. In the third pattern (the CP-peak type), warm SST anomalies are centered in the Niño-4 region and can extend as far west as 160°E, with very weak SST anomalies in the eastern Pacific (Fig. 4e).

The corresponding BMN series (second row in Fig. 5) shows that the frequency of EP-like-peak types has decreased significantly in the past 30 years (six before 1978 and three after 1978) whereas CP-peak types have occurred more frequently since 1990, which is consistent with some previous studies (Ashok et al. 2007; Kao and Yu 2009; Kug et al. 2009; Lee and McPhaden 2010). The EP-peak type only appears three times (1972/73, 1982/83, and 1997/98), but these events have drawn considerable attention. Hong et al. (2014), for example, referred to them as “super El Niño” events and showed that the SSH difference between the eastern and western Pacific, and the zonal wind stress in the central equatorial Pacific, is significantly higher than in other events. Cai et al. (2014) classified El Niño events into extreme El Niño and moderate El Niño, with the former type being identical to the EP-peak type in this study. They also pointed out that the two types can be distinguished by their zonal and meridional SST gradient as well as rainfall in the Niño-3 region.

Figures 4f and 4g show that SST anomaly patterns during the end phase of El Niño can be classified into EP-end and CP-end types. In the former, residual positive SST anomalies are mainly distributed in the eastern Pacific, especially off the Peruvian coast (Fig. 4f), whereas in the latter they remain in the Niño-4 region (Fig. 4g). The BMN series (third row of Fig. 5) shows that more than two-thirds of El Niño events ended in CP-end type over the past six decades. However, the occurrence of the two types shows an interdecadal change: the CP-end type clearly dominates in the 1960s, early 1970s, and 2000s, whereas the EP-end type dominates in the 1950s and 1970s.

c. La Niña phase types

Based on the location of minimum SST anomalies, the distribution of SST anomalies during the onset phase of La Niña can be classified into EP-onset types (Fig. 6a) and CP-onset types (Fig. 6b). In the EP-onset type, cold water first builds in the eastern equatorial Pacific, especially off the Peruvian coast (Fig. 6a), and appears simultaneously in the central equatorial Pacific and “interdecadal pathway” (Zhou et al. 2007) region. In the CP-onset type, strong positive SST anomalies occur in the southeastern equatorial Pacific (Fig. 6b). The corresponding BMN series (fourth row of Fig. 5) indicates that in the last six decades, most La Niña events are EP-onset type while only three events are CP-onset type. Note that all the three CP-onset La Niña events appear after the 1980s. These results indicate that La Niña mainly stems from the eastern equatorial Pacific, especially before the 1980s.

Fig. 6.

As in Fig. 4, but for La Niña.

Fig. 6.

As in Fig. 4, but for La Niña.

Although SST anomaly patterns in the peak phase of La Niña are also classified into three types (Figs. 6c–e), their shape and strength are very different from El Niño. In the first pattern, negative SST anomalies mainly cover the eastern equatorial Pacific and are centered in the Niño-3 region (Fig. 6c), so it can be called an EP-peak type. These events are very different from the canonical EP-peak type, but have a similar shape to the EP-like-peak type of El Niño (Fig. 4d). In the third pattern, cold SST anomalies are located in the central Pacific with maximum anomalies in the Niño-4 region, so it can be named a CP-peak type. There are two cold tongues associated with the CP-peak type, one extending to the interdecadal pathway region and another extending to the Peruvian coast (Fig. 6e). The second pattern is called the mixed (MIX)-peak type, as it has features of both EP- and CP-peak types. In this pattern, negative SST anomalies are strong in both Niño-3 and Niño-4 regions, with their primary center in the Niño-3.4 region and secondary center near the west coast of Mexico (Fig. 6d). The spatial shape of the MIX-peak type is clearly distinguishable from any El Niño peak type. Moreover, it should be noted that the EP-peak type is the strongest El Niño type, while the MIX-peak type has the largest SST anomaly amplitude of all La Niña types. These differences reflect the spatial asymmetry between El Niño and La Niña.

The BMN series (fifth row in Fig. 5) shows that 45% La Niña events are MIX-peak type, 35% are EP-peak type, and only 20% are CP-peak type. The CP-peak type often occurs either as the secondary peak in an extended La Niña event with multiple peaks or in a weak La Niña event. Using the primary peak of each event in FDR–SOM analysis produces only two statistically different patterns (MIX and EP; figure not shown). Thus, unlike El Niño, the CP-peak type is not a significant component of La Niña events.

The spatial distribution of SST anomalies in the end phase of La Niña can be classified into EP-end and CP-end types. The EP-end-type La Niña (Fig. 6f) is associated with negative SST anomalies in the eastern Pacific centered in the eastern Niño-3 region. For the CP-end type (Fig. 6g), residual cold SST anomalies mainly appear in the equatorial central Pacific and interdecadal pathway region, with one center located in the Niño-3.4 region and another near the west coast of Mexico. The spatial patterns of EP- and CP-end-type La Niña events are therefore different from those associated with El Niño events. The BMN series (bottom row in Fig. 5) shows that the EP-end type occurred during 60% of La Niña events over the past six decades, and there is no significant trend. However, the CP-end type shows multidecadal variability, and only occurs in the 1970s and 2000s.

d. General types of El Niño and La Niña

To compare our results with previous studies, each El Niño or La Niña must be classified into a general type. The general type of each event is mainly dependent on its peak type, with its onset type and end type taken into consideration. If an El Niño has an EP or EP-like type peak and begins or ends in an EP type, it can be treated as general EP-type El Niño (hereinafter EP El Niño). Similarly, if an El Niño has a CP-type peak and begins or ends in a CP type, it is classified as a general CP type (hereinafter CP El Niño). A total of 8 EP El Niño events and 9 CP El Niño events are identified from the total of 19 El Niño events. The remaining 2 events, which both begin and end in a CP type but have EP-like peaks, are excluded from the analysis because of the difficulty comparing them with previous studies.

Out of 15 La Niña events, 6 events have EP-type peaks, 7 have MIX-type peaks, and only 2 events have CP-type peaks. For the two events with CP-type peaks, one has EP-type onset and end phases, while the other both begins and terminates in a CP-type pattern. Therefore, La Niña CP types are also excluded from the analysis because they do not have a consistent classification. For all La Niña events with EP-type peaks, there is only one event with a CP-type onset, while all others both begin and end in EP-type. All such events are treated as general EP-type La Niña (hereinafter EP La Niña). The remaining events are all classified as general MIX type (hereinafter MIX La Niña), as they all have MIX-type peak phases with stronger SST anomalies than the other types. Using this classification, a total of six EP La Niña and seven MIX La Niña events are identified.

The composite evolutions of the horizontal distribution of SST anomalies for the EP and CP El Niño are shown in Fig. 7 alongside corresponding SSH and wind stress anomalies that pass the 95% significance test. For the wind stress vector, the significance test is applied to both zonal and meridional components, and the wind stress is treated as significant if either component passes the test.

Fig. 7.

The composites of anomalies of SST (color; °C), SSH (contour, interval is 0.025 m), and wind stress (vector; N m−2) for the (left) general EP El Niño and (right) CP El Niño from month −12 to 12. Month 0 represents the time of the peak phase. Results passing the significance test at the 95% confidence level are presented for SST and wind stress, stippled for SSH. Dashed contours are for negative values; zero contours are omitted.

Fig. 7.

The composites of anomalies of SST (color; °C), SSH (contour, interval is 0.025 m), and wind stress (vector; N m−2) for the (left) general EP El Niño and (right) CP El Niño from month −12 to 12. Month 0 represents the time of the peak phase. Results passing the significance test at the 95% confidence level are presented for SST and wind stress, stippled for SSH. Dashed contours are for negative values; zero contours are omitted.

A few months before the onset of EP El Niño, negative SST anomalies tend to occur in the central and eastern Pacific with a La Niña–like pattern. Meanwhile, the thermocline tends to deepen in the western Pacific resulting from increased easterly trade winds (Fig. 7a). This result suggests that onset of an EP El Niño event often follows the termination of a La Niña event. In an EP El Niño, positive SST anomalies first appear off the coast of South America (Fig. 7b) and then propagate westward to the central Pacific (Fig. 7c). At the same time, the thermocline deepens in the eastern Pacific, while shoaling occurs in the western Pacific. The shoaling signal then propagates eastward along the equator (Figs. 7c–g). During the peak phase, positive SST anomalies are centered over the eastern Pacific near the Peruvian coast (Fig. 7d). Positive SST anomalies also disappear over the central Pacific or eastern Pacific during the EP El Niño decay phase (Fig. 7f). Cold SST anomalies may reemerge at the equator when the shoaling of the thermocline reaches the eastern Pacific (Fig. 7g).

Before the onset of a CP El Niño, there are no significant cold SST anomalies in the central and eastern Pacific (Fig. 7h), indicating that these events may be initiated independently. Warm SST anomalies generally stem from the central Pacific (Fig. 7i) and then extend both eastward and westward (Fig. 7j). During the development of a CP El Niño, a deepening of the thermocline also occurs in the central Pacific and propagates eastward (Figs. 7i–k). In the peak phase, strong positive SST anomalies in the Niño-4 region enhance easterly trade winds in the eastern Pacific (Fig. 7k), which may help to lift the thermocline and increase upwelling off the coast. Decreased SST in the far eastern Pacific further boosts trade winds, in turn accelerating cooling of ocean surface and shoaling of the thermocline off the coast (Figs. 7l,m). Enhanced trade winds may also transport a large amount of cold water from the cold tongue to the central Pacific, which would eventually lead to the disappearance of positive SST anomalies (Fig. 7m).

These results suggest that the two types of El Niño have many differences in SST patterns and wind stress distributions, as well as thermocline structure. Furthermore, EP El Niño events are more likely to produce transitions from or to La Niña. On the other hand, CP El Niño events tend to have little correlation with preceding or following events. These findings are consistent with Kao and Yu (2009) to some degree.

The evolution of EP and MIX La Niña are shown in Fig. 8. Before the onset of EP La Niña, there are westerly anomalies on the western equatorial Pacific, shoaling the thermocline in the far western Pacific (Fig. 8b). This shoaling signal seems to propagate eastward along the equator and help to generate the negative SST anomalies in the Niño-3 region (Fig. 8c). In contrast, before the onset of a MIX La Niña, strong northwesterly anomalies occur in the northern tropical Pacific, which may help to lift the thermocline and reduce SST in the eastern Pacific. Thus, negative SST anomalies often stem from the coast of South America (Fig. 8h).

Fig. 8.

As in Fig. 7, but for (left) general EP La Niña and (right) MIX La Niña, respectively.

Fig. 8.

As in Fig. 7, but for (left) general EP La Niña and (right) MIX La Niña, respectively.

Negative SST anomalies extend both westward and eastward after the onset of an EP La Niña. They tend to be centered around 120°W in the peak phase (Fig. 8d) and then decay rapidly toward both the east and the west (Fig. 8e). Weak SST anomalies combined with subtle thermocline and wind stress variations are not expected to promote strong air–sea or thermocline feedbacks. Thus, EP La Niña events are usually very weak and unlikely to induce an El Niño.

In contrast, negative SST anomalies propagate west and develop rapidly in Niño-3 and then in Niño-4 regions after the onset of MIX La Niña (Figs. 8i,j). During the developing period, intensified easterly trade winds in the western and central Pacific help to deepen the thermocline in the far western Pacific, while lifting the thermocline in the central and eastern Pacific, and producing positive air–sea feedback. In the peak phase, strong negative SST anomalies tend to be centered in the Niño-3.4 region, inducing strong easterly anomalies in the western and central Pacific, and westerly anomalies in the eastern Pacific (Fig. 8k). Weakened trade winds would suppress any upwelling, so negative SST anomalies first decrease in the eastern Pacific and then decay slowly in the equatorial central Pacific (Figs. 8l,m). However, southwesterly anomalies often increase in the southeastern Pacific during the decaying phase (Fig. 8m), enhancing the upwelling in the equatorial eastern Pacific. Thus, negative SST anomalies can persist for a long time and even develop again (Fig. 8n), indicating that there may be a resurgence of MIX La Niña events, or that they may have multiple peaks over a long time period.

These results indicate that MIX La Niña is very different from EP La Niña. The SST anomalies in the former are usually much stronger and more persistent than the latter. In addition, zonal and meridional gradient of thermocline depth as well as wind stress anomalies in the former are much larger than the latter. Thus, MIX La Niña events seem more significant than EP La Niña events.

e. The relationships between the types of contiguous El Niño and La Niña

In addition to the identification of different types of El Niño and La Niña in different phase, it is also useful to understand which type of La Niña (El Niño) is likely to follow a certain type of El Niño (La Niña). The relationship between contiguous El Niño and La Niña may not only help us to understand inter-ENSO variability, but may also contribute to the prediction of ENSO. In this study, “contiguous” is defined as a time interval of less than 6 months between the end of an El Niño and the onset of a following La Niña (hereinafter a positive transition), or the end of a La Niña and the onset of a following El Niño (hereinafter a negative transition). Based on this criterion, there are a total of 12 positive transitions and 8 negative transitions during 1951–2011. Thus, after their rapid decay, El Niño events are more likely to induce a La Niña event rather than the reverse scenario (i.e., La Niña inducing an El Niño). Note that all EP El Niño excepting the 1982/83 event are included in the positive transitions and all MIX La Niña are included in the negative transitions. This indicates that EP El Niño events are often induced by La Niña and MIX La Niña events are always induced by El Niño.

In positive transitions, three EP El Niño events and three CP El Niño events transition to MIX La Niña events, while one EP El Niño and three CP El Niño events transition to EP La Niña. In this regard, both EP and CP El Niño events can induce a strong La Niña, which is different from the results of Yu et al. (2011). However, EP El Niño is indeed more likely to transition to MIX La Niña than to EP La Niña, which is in some degree consistent with their conclusions. In negative transitions, all MIX La Niña events transition to EP El Niño and most EP La Niña events also transition to EP El Niño. Thus, there are no robust connections between the general types of El Niño and La Niña, except that a MIX La Niña is almost impossible to induce a CP El Niño and the La Niña events are more likely to induce the EP El Niño instead of CP El Niño.

The end type of El Niño or La Niña and the onset type of the succeeding La Niña or El Niño are further investigated in order to describe the relationships between the types of contiguous El Niño and La Niña in more detail. However, in positive transitions, all EP-end-type El Niño events transition to CP-onset-type La Niña events, and all CP-end-type El Niño events transition to EP-onset-type La Niña events. This suggests that an EP (CP)-end-type El Niño event is very likely to induce a CP (EP)-onset-type La Niña. That is to say, the end pattern of an El Niño can strongly influence the onset pattern of the following La Niña. In addition, it is found that most El Niño in the positive transitions are CP-end type, indicating that an El Niño terminating in the central equatorial Pacific is more likely to induce a La Niña. Similarly, in the negative transitions, all CP-end-type La Niña events can transition to EP-onset-type El Niño events. However, most EP-end-type La Niña events can also transition to EP-onset-type El Niño events. Further examination during 1951–2011 shows that all EP-onset-type El Niño events have transitioned from La Niña, whereas only one CP-onset-type El Niño has transitioned from La Niña. These results indicate that EP-onset-type El Niño events are always induced by a previous La Niña, whereas CP El Niño events are more independent of previous SST anomalies.

To better understand the atmosphere–ocean dynamics associated with these transitions, we explore Hovmöller diagrams of anomalous SST, SSH (which is highly correlated with thermocline depth), and surface wind stress vectors composited during different El Niño and La Niña transition scenarios (Fig. 9). A transition from a CP-end-type El Niño to an EP-onset-type La Niña is shown in Fig. 9a. Significant positive SST anomalies, a deepened thermocline, and strong westerly anomalies occur over the central Pacific a few months before the termination of El Niño. Warm SST anomalies favor stronger atmospheric deep convection, inducing the westerly anomalies in the western–central Pacific. Such westerly anomalies lead to thermocline shoaling in the western equatorial Pacific. Coupled air–sea interactions then cause the shoaling signal to gradually propagate toward the east, following the behavior of a slow SST mode (Neelin 1991; Wang and Weisberg 1994). Once the shoaling signal arrives in the eastern Pacific, it lifts the thermocline. More importantly, the westward shift of the SST anomaly center leads to a climatology-like thermocline in the eastern Pacific and a westward shift of the deep convection center. The latter intensifies easterly trade winds in the eastern Pacific and further lifts the thermocline, bringing a cold upwelling from the subsurface to the shallow upper ocean. Negative SST anomalies may inhibit atmospheric convection (figure not shown) and thus reinforce easterly winds to the west of the negative SST anomalies. Therefore, a positive air–sea feedback may be activated, ocean surface cooling is accelerated, and a La Niña would grow rapidly in the eastern Pacific.

Fig. 9.

Composite Hovmöller diagrams of the anomalies of SST (color; °C), SSH (contour, interval is 0.02 m), and wind stress (vector; N m−2) averaged between 5°S and 5°N in four different transition scenarios in ENSO cycle: (a) positive transition from the El Niño ending in CP-end type; (b) positive transition from the El Niño ending in EP-end type; (c) negative transition from La Niña ending in CP-end type; and (d) El Niño initiated in EP-onset type. Month 0 represent the time of the end phase of El Niño for (a) and (b), the end phase of La Niña for (c), and the onset phase of El Niño for (d). Results passing the significance test at the 95% confidence level are presented for SST and wind stress, stippled for SSH. Dashed contours are for negative values; zero contours are omitted.

Fig. 9.

Composite Hovmöller diagrams of the anomalies of SST (color; °C), SSH (contour, interval is 0.02 m), and wind stress (vector; N m−2) averaged between 5°S and 5°N in four different transition scenarios in ENSO cycle: (a) positive transition from the El Niño ending in CP-end type; (b) positive transition from the El Niño ending in EP-end type; (c) negative transition from La Niña ending in CP-end type; and (d) El Niño initiated in EP-onset type. Month 0 represent the time of the end phase of El Niño for (a) and (b), the end phase of La Niña for (c), and the onset phase of El Niño for (d). Results passing the significance test at the 95% confidence level are presented for SST and wind stress, stippled for SSH. Dashed contours are for negative values; zero contours are omitted.

Figure 9b illustrates the transition from an EP-end-type El Niño to a CP-onset-type La Niña. An important feature of this transition is that extremely strong SST anomalies and a deep thermocline in the eastern Pacific occur well before the termination of El Niño. Strong SST anomalies sustain abnormally deep convection in the eastern Pacific, inducing significant westerly anomalies in the central and eastern Pacific. This longitudinal distribution of westerly anomalies produces a massive shoaling of the thermocline in the western Pacific. This shoaling then penetrates toward the east in accordance with the slow SST mode. Meanwhile, with the eastward shift of SST anomalies, deep convection weakens in the central Pacific but strengthens in the eastern Pacific, causing westerly anomalies to shift to the eastern Pacific. The eastward propagation of elevated thermocline anomalies, combined with westerly anomalies in the eastern Pacific, accelerates thermocline shoaling and ocean surface cooling in the central Pacific, thereby activating a positive atmosphere–ocean feedback and rapid La Niña generation in the central Pacific.

The two scenarios discussed above are very similar to the composite mean (CM)-EOF1 [or CM + second rotated EOF (REOF2)] and CM+EOF2 (or CM+REOF1) in Lee et al. (2014a), respectively. The present study describes how a CP (EP)-end-type El Niño induces an EP (CP)-onset-type La Niña, while Lee et al. (2014a) explain why a weak or early-onset El Niño favors a transition to La Niña. Indeed, the CM+REOF1 also shows that a CP-end-type El Niño tends to transition to an EP-onset-type La Niña, while the CM+REOF2 shows that an EP-end-type El Niño tends to transition to a CP-onset-type La Niña. Although the purpose and methods of the two studies are different, the results and key physical processes described are consistent, and both investigations provide a fresh perspective on the main features of El Niño–La Niña transitions.

The transition from a CP-end-type La Niña to an EP-onset-type El Niño is illustrated in Fig. 9c. Strong negative SST anomalies are centered in the central Pacific well before the termination of La Niña. This SST anomaly distribution increases (decreases) the SST gradient and thermocline between the western and central Pacific (eastern and central Pacific). This process intensifies easterly trade winds in the central and western equatorial Pacific, and weakens those in the eastern Pacific. The anomalously strong trade winds deepen the thermocline in the western Pacific, and the deepening slowly penetrates eastward. Meanwhile, weakening trade winds in the eastern Pacific inhibit coastal upwelling, leading to increased SST. On the other hand, increased solar radiation in the eastern Pacific from weakening of deep convection would also accelerate the heating on the thin mixed layer. SST therefore increases further with the arrival of deepened thermocline anomalies, which may trigger an El Niño in the eastern Pacific, especially off the Peruvian coast.

The composite of all EP-onset-type El Niño events is shown in Fig. 9d. Before the EP-onset-type El Niño, the SST anomalies in the equatorial Pacific show a La Niña pattern. This pattern is associated with negative SST anomalies and elevated thermocline anomalies in the central–eastern Pacific, and positive SST anomalies and deepened thermocline anomalies in the western Pacific. Because of the increased zonal SST gradient and westward shift of the deep convection center, easterly trade winds are intensified in the western and central Pacific, and thus may further deepen the thermocline in the western Pacific. The eastward propagation of deepened thermocline anomalies contributes to a deepening thermocline in the eastern Pacific. On the other hand, as the zonal gradient of SSH is much stronger than the climatological mean, this may strengthen the equatorial undercurrent and transport warm water eastward to the eastern Pacific. These two processes, combined with increased solar radiation, lead to warming of the mixed layer in the far eastern Pacific and increased westerly winds to the west of the positive SST anomalies. This appears to activate a Bjerknes feedback (Bjerknes 1969), leading to an EP-onset-type El Niño.

4. Discussion and conclusions

Spatial patterns of SST anomalies are identified for all El Niño and La Niña events in their onset, peak, and end phases using a new definition of El Niño and La Niña and combined SOM and FDR methods. El Niño and La Niña events are further classified into general types based on their onset, peak, and end phases types, and especially the peak phase type. The horizontal distributions of SST, SSH, and wind stress anomalies are compared during the life cycle of different El Niño and La Niña general types, and the relationship between the types of the contiguous El Niño and La Niña are investigated.

It is found that SST patterns can be classified into EP and CP types for both the onset and end phases of El Niño and La Niña. El Niño events tend to end in a CP type, whereas La Niña events are more likely to begin in an EP type. There are three types of SST anomaly patterns for the peak phase of both El Niño and La Niña events, but they have significantly different spatial features and strengths. El Niño peak phases include EP, EP-like, and CP types, with the first two both classified into EP type in other studies (Kao and Yu 2009; Yeh et al. 2009; Yu and Kim 2010). La Niña peak phases include EP, CP, and MIX types with both CP and MIX likely viewed as CP type in previous studies (Kao and Yu 2009; Yu et al. 2011; Shinoda et al. 2011; Yuan and Yan 2013; Zhang et al. 2015). Indeed, both EP-peak-type El Niño and CP-peak-type La Niña occur infrequently. It is interesting that there is some decadal variability in the occurrence frequencies of different event types. El Niño events, for example, are more likely to have EP onset before 1978 but likely have CP onset after 1978. The number of EP-like-peak El Niño events decreased significantly in the past 30 years, while CP-peak El Niño events increased after 1990. In the 1950s and during 1975–85, EP-end El Niño events occurred more frequently, while CP-end El Niño events dominated in the 1960s, early 1970s, and 2000s. All La Niña events have EP onset before 1980, whereas only three events have CP onset after 1980. In the periods 1950–70 and 1980–2000, all La Niña events are of EP-end type, whereas most La Niña terminated in CP type in the 1970s and 2000s. As suggested by Chung and Li (2013) and Xiang et al. (2013), these decadal changes are probably caused by a decadal mean state change.

According to the onset, peak, and end phases types, El Niño events can be classified into general EP and CP types and La Niña events can be classified into general EP and MIX types. There is no robust connection between the general types of contiguous El Niño and La Niña events. However, it should be noted that MIX La Niña can almost never induce a CP El Niño, whereas an EP El Niño is generally induced by a La Niña.

However, there are strong relationships between the end types of El Niño and the onset types of subsequent La Niña. The EP-end El Niño events always transition to CP-onset La Niña events, and CP-end El Niño events always transition to EP-onset La Niña events. In the former situation, enhanced easterlies converging from the east toward the central Pacific before and during the end phase of El Niño presumably induce a positive air–sea feedback to produce and amplify negative SST anomalies in the eastern Pacific. In the latter situation, the eastward propagation of elevated thermocline anomalies due to a slow SST mode plays a crucial role, combined with the development of westerly anomalies in the eastern Pacific. On the other hand, relationships between the end types of La Niña and the onset types of contiguous El Niño events are not so robust. Although CP-end La Niña events generally transition to EP-onset-type El Niño events, most EP-end La Niña events can also transition to EP-onset El Niño events. Furthermore, the EP-onset El Niño events are always induced by a previous La Niña. The key processes underlying the transition from a CP-end-type La Niña to an EP-onset-type El Niño are weakening trade winds in the eastern Pacific before and during the end phase of La Niña, and the consequent development of coupled Bjerknes feedbacks. Before all EP-onset El Niño events, the previous La Niña events provide favorable conditions for the development of positive SST anomalies off the Peruvian coast, such as westerly anomalies in the far eastern Pacific, enhanced equatorial undercurrents, and the eastward propagation of deepened thermocline anomalies.

To validate our results, we performed the SOM–FDR procedure again using another dataset (HadISST). It is found that both datasets produce similar patterns and a consistent BMN series, aside from differences in the classification of some complicated events. For example, the peak phase of 1991/92 El Niño is classified as CP type by ERSST but as EP-like type by HadISST, and the peak phase of 2006/07 El Niño is identified as EP-like type in ERSST but as CP type in HadISST. We also used some indices from previous studies, such as in Yeh et al. (2009) and Ren and Jin (2011), to classify ENSO events. While the results (not shown) differ to some extent from those obtained by SOM, they are consistent in many cases, especially for the classification of El Niño end phase types and the La Niña onset phase types. Thus, the relationships between end types of El Niño and onset types of subsequent La Niña are not affected by the index used in the analysis. The 1953/54, 1963/64, and 1982/83 El Niño events were classified as CP-onset type by SOM but EP-onset type using the index from previous studies. Indeed, the evolution of SST anomalies shows that they did not initiate from the Peruvian coast. Thus, SOM clustering appears to be more accurate and robust than previous methods.

In conclusion, the SOM–FDR method used in this study is effective in identifying all El Niño and La Niña phase types. As in any clustering method, there is still some uncertainty in the results, but the approach seems to produce accurate and reliable results. Our results confirm some known features of different El Niño types, but also reveal more about the detailed characteristics of the three phases. Moreover, the results suggest that there are generally two types of La Niña, which differ in amplitude, period, and the horizontal distribution of SST, and show spatial asymmetry with El Niño. More importantly, our results give new insight into the relationships and dynamics between the continuous El Niño and La Niña. This provides a possible predictor for the transitions between El Niño and La Niña. Future work will focus on further exploration of atmosphere–ocean dynamics involving different ENSO onset, peak, and end types to understand more about inter–El Niño (or inter–La Niña) variability and mechanisms for pattern asymmetry between El Niño and La Niña.

Acknowledgments

Three anonymous reviewers provided careful comments on the submitted manuscript, which helped improve this article. The authors thank Prof. Nathaniel C. Johnson at the University of Hawaii for providing valuable suggestions and methods. This work was supported by the National Basic Research Program of China (2013CB956203 and 2015CB453200) and National Natural Science Foundation of China (41490642, 41575062, and 41520104008).

REFERENCES

REFERENCES
Aceituno
,
P.
,
1992
:
El Niño, the Southern Oscillation, and ENSO: Confusing names for a complex ocean–atmosphere interaction
.
Bull. Amer. Meteor. Soc.
,
73
,
483
485
.
Ashok
,
K.
,
S. K.
Behera
,
S. A.
Rao
,
H. Y.
Weng
, and
T.
Yamagata
,
2007
:
El Niño Modoki and its possible teleconnection
.
J. Geophys. Res.
,
112
,
C11007
, doi:.
Banholzer
,
S.
, and
S.
Donner
,
2014
:
The influence of different El Niño types on global average temperature
.
Geophys. Res. Lett.
,
41
,
2093
2099
, doi:.
Battisti
,
D. S.
, and
A. C.
Hirst
,
1989
:
Interannual variability in a tropical atmosphere–ocean model: Influence of the basic state, ocean geometry and nonlinearity
.
J. Atmos. Sci.
,
46
,
1687
1712
, doi:.
Bjerknes
,
J.
,
1969
:
Atmospheric teleconnections from the equatorial Pacific
.
Mon. Wea. Rev.
,
97
,
163
172
, doi:.
Cai
,
W.
, and
T.
Cowan
,
2009
:
La Niña Modoki impacts Australia autumn rainfall variability
.
Geophys. Res. Lett.
,
36
,
L12805
, doi:.
Cai
,
W.
, and Coauthors
,
2014
:
Increasing frequency of extreme El Niño events due to greenhouse warming
.
Nat. Climate Change
,
4
,
111
116
, doi:.
Cassano
,
E. N.
,
A. H.
Lynch
,
J. J.
Cassano
, and
M. R.
Koslow
,
2006
:
Classification of synoptic patterns in the western Arctic associated with extreme events at Barrow, Alaska, USA
.
Climate Res.
,
30
,
83
97
, doi:.
Chao
,
J.
,
S.
Yuan
,
Q.
Chao
, and
J.
Tian
,
2002
:
A data analysis study on the evolution of the EI Niño/La Niña cycle
.
Adv. Atmos. Sci.
,
19
,
837
844
, doi:.
Chattopadhyay
,
R.
,
A. K.
Sahai
, and
B.
Goswami
,
2008
:
Objective identification of nonlinear convectively coupled phases of monsoon intraseasonal oscillation: Implications for prediction
.
J. Atmos. Sci.
,
65
,
1549
1569
, doi:.
Chattopadhyay
,
R.
,
A.
Vintzileos
, and
C.
Zhang
,
2013
:
A description of the Madden–Julian oscillation based on a self-organizing map
.
J. Climate
,
26
,
1716
1732
, doi:.
Chung
,
P.-H.
, and
T.
Li
,
2013
:
Interdecadal relationship between the mean state and El Niño types
.
J. Climate
,
26
,
361
379
, doi:.
Fu
,
C. B.
,
1985
:
Two sea surface temperature warming patterns during El Niño
.
Chin. Sci. Bull.
,
30
,
596
599
.
Giese
,
B. S.
, and
S.
Ray
,
2011
:
El Niño variability in simple ocean data assimilation (SODA), 1871–2008
.
J. Geophys. Res.
,
116
,
C02024
, doi:.
Gleckler
,
P. J.
, and Coauthors
,
2012
:
Human-induced global ocean warming on multidecadal timescales
.
Nat. Climate Change
,
2
,
524
529
, doi:.
Hansen
,
J.
,
M.
Sato
,
R.
Ruedy
,
K.
Lo
,
D. W.
Lea
, and
M.
Medina-Elizade
,
2006
:
Global temperature change
.
Proc. Natl. Acad. Sci. USA
,
103
,
14 288
14 293
, doi:.
Hong
,
L.-C.
,
LinHo
, and
F.-F.
Jin
,
2014
:
A Southern Hemisphere booster of super El Niño
.
Geophys. Res. Lett.
,
41
,
2142
2149
, doi:.
Hu
,
Z.-Z.
,
A.
Kumar
,
B.
Jha
,
W.
Wang
,
B.
Huang
, and
B.
Huang
,
2012
:
An analysis of warm pool and cold tongue El Niños: Air–sea coupling processes, global influences, and recent trends
.
Climate Dyn.
,
38
,
2017
2035
, doi:.
Iskandar
,
I.
,
2009
:
Variability of satellite-observed sea surface height in the tropical Indian Ocean: Comparison of EOF and SOM analysis
.
Makara Seri Sains
,
13
(
2
),
173
179
, doi:.
Iskandar
,
I.
,
2010
:
Seasonal and interannual patterns of sea surface temperature in Banda Sea as revealed by self-organizing maps
.
Cont. Shelf Res.
,
30
,
1136
1148
, doi:.
Jin
,
B.
,
G.
Wang
,
Y.
Liu
, and
R.
Zhang
,
2010
:
Interaction between the East China Sea Kuroshio and the Ryukyu Current as revealed by the self-organizing map
.
J. Geophys. Res.
,
115
,
C12047
, doi:.
Jin
,
F.-F.
,
1997a
:
An equatorial ocean recharge paradigm for ENSO. Part I: Conceptual model
.
J. Atmos. Sci.
,
54
,
811
829
, doi:.
Jin
,
F.-F.
,
1997b
:
An equatorial ocean recharge paradigm for ENSO. Part II: A stripped-down coupled model
.
J. Atmos. Sci.
,
54
,
830
847
, doi:.
Johnson
,
N. C.
,
2013
:
How many ENSO flavors can we distinguish?
J. Climate
,
26
,
4816
4827
, doi:.
Johnson
,
N. C.
, and
S. B.
Feldstein
,
2010
:
The continuum of North Pacific sea level pressure patterns: Intraseasonal, interannual, and interdecadal variability
.
J. Climate
,
23
,
851
867
, doi:.
Johnson
,
N. C.
,
S. B.
Feldstein
, and
B.
Tremblay
,
2008
:
The continuum of Northern Hemisphere teleconnection patterns and a description of the NAO shift with the use of self-organizing maps
.
J. Climate
,
21
,
6354
6371
, doi:.
Kao
,
H. Y.
, and
J. Y.
Yu
,
2009
:
Contrasting eastern-Pacific and central-Pacific types of ENSO
.
J. Climate
,
22
,
615
632
, doi:.
Kim
,
H. M.
,
P. J.
Webster
, and
J. A.
Curry
,
2009
:
Impact of shifting patterns of Pacific Ocean warming on North Atlantic tropical cyclones
.
Science
,
325
,
77
–80, doi:.
Kohonen
,
T.
,
2001
: Self-Organizing Maps. 3rd ed. Springer, 501 pp.
Kohonen
,
T.
,
J.
Hynninen
,
J.
Kangas
, and
J.
Laaksonen
,
1995
: SOM_PAK: The self- organizing maps program package. Helsinki University of Technology, 27 pp. [Available online at http://www.cis.hut.fi/research/som_pak/som_doc.txt.]
Kug
,
J.-S.
, and
Y.-G.
Ham
,
2011
:
Are there two types of La Niña?
Geophys. Res. Lett.
,
38
,
L16704
, doi:.
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
, doi:.
Kug
,
J.-S.
,
J.
Choi
,
S.-I.
An
,
F.-F.
Jin
, and
A. T.
Wittenberg
,
2010
:
Warm pool and cold tongue El Niño events as simulated by the GFDL 2.1 coupled GCM
.
J. Climate
,
23
,
1226
1239
, doi:.
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:.
Larson
,
S.
,
S.-K.
Lee
,
C.
Wang
,
E.-S.
Chung
, and
D.
Enfield
,
2012
:
Impacts of non-canonical El Niño patterns on Atlantic hurricane activity
.
Geophys. Res. Lett.
,
39
,
L14706
, doi:.
Lee
,
S.-K.
,
P. N.
DiNezio
,
E.-S.
Chung
,
S.-W.
Yeh
,
A. T.
Wittenberg
, and
C.
Wang
,
2014a
:
Spring persistence, transition and resurgence of El Niño
.
Geophys. Res. Lett.
,
41
,
8578
8585
, doi:.
Lee
,
S.-K.
,
B. E.
Mapes
,
C.
Wang
,
D. B.
Enfield
, and
S. J.
Weaver
,
2014b
:
Springtime ENSO phase evolution and its relation to rainfall in the continental U.S
.
Geophys. Res. Lett.
,
41
,
1673
1680
, doi:.
Lee
,
T.
, and
M. J.
McPhaden
,
2010
:
Increasing intensity of El Niño in the central- equatorial Pacific
.
Geophys. Res. Lett.
,
37
,
L14603
, doi:.
Leloup
,
J.
,
M.
Lengaigne
, and
J. P.
Boulanger
,
2008
:
Twentieth century ENSO characteristics in the IPCC database
.
Climate Dyn.
,
30
,
277
291
, doi:.
Levitus
,
S.
,
J. I.
Antonov
,
T. P.
Boyer
, and
C.
Stephens
,
2000
:
Warming of the World Ocean
.
Science
,
287
,
2225
2229
, doi:.
Levitus
,
S.
, and Coauthors
,
2012
:
World Ocean heat content and thermosteric sea level change (0–2000 m), 1955–2010
.
Geophys. Res. Lett.
,
39
,
L10603
, doi:.
L’Heureux
,
M. L.
,
D. C.
Collins
, and
Z.-Z.
Hu
,
2013
:
Linear trends in sea surface temperature of the tropical Pacific Ocean and implications for the El Niño–Southern Oscillation
.
Climate Dyn.
,
40
,
1223
1236
, doi:.
Li
,
C. Y.
, and
M. Q.
Mu
,
1999
:
El Niño occurrence and sub-surface ocean temperature anomalies in the Pacific warm pool
.
Chin. J. Atmos. Sci.
,
5
,
513
521
.
Li
,
C. Y.
, and
M. Q.
Mu
,
2002
:
A further study of the essence of ENSO
.
Chin. J. Atmos. Sci.
,
4
,
309
328
.
Li
,
T.
,
1997
:
Phase transition of the El Niño–Southern Oscillation: A stationary SST mode
.
J. Atmos. Sci.
,
54
,
2872
2887
, doi:.
Li
,
X.
, and
C.
Li
,
2014
:
Occurrence of two types of El Niño events and the subsurface ocean temperature anomalies in the equatorial Pacific
.
Chin. Sci. Bull.
,
59
,
3471
3483
, doi:.
Lin
,
G.-F.
, and
L.-H.
Chen
,
2006
:
Identification of homogeneous regions for regional frequency analysis using the self-organizing map
.
J. Hydrol.
,
324
,
1
9
, doi:.
Liu
,
Y.
, and
R. H.
Weisberg
,
2005
:
Patterns of ocean current variability on the West Florida Shelf using the self-organizing maps
.
J. Geophys. Res.
,
110
,
C06003
, doi:.
Liu
,
Y.
,
R. H.
Weisberg
, and
L. K.
Shay
,
2007
:
Current patterns on the West Florida Shelf from joint self-organizing map analyses of HF radar and ADCP data
.
J. Atmos. Oceanic Technol.
,
24
,
702
712
, doi:.
Neelin
,
J. D.
,
1991
:
The slow sea surface temperature mode and the fast-wave limit: Analytic theory for tropical interannual oscillation and experiments in a hybrid coupled model
.
J. Atmos. Sci.
,
48
,
584
606
, doi:.
Rayner
,
N. A.
,
D. E.
Parker
,
E. B.
Horton
,
C. K.
Folland
,
L. V.
Alexander
,
D. P.
Rowell
,
E. C.
Kent
, and
A.
Kaplan
,
2003
:
Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century
.
J. Geophys. Res.
,
108
,
4407
, doi:.
Ren
,
H. L.
, and
F.-F.
Jin
,
2011
:
Niño indices for two types of ENSO
.
Geophys. Res. Lett.
,
38
,
L04704
, doi:.
Ren
,
H. L.
, and
F.-F.
Jin
,
2013
:
Recharge oscillator mechanisms in two types of ENSO
.
J. Climate
,
26
,
6506
6523
, doi:.
Schopf
,
P. S.
, and
M. J.
Suarez
,
1988
:
Vacillations in a coupled ocean–atmosphere model
.
J. Atmos. Sci.
,
45
,
549
566
, doi:.
Shinoda
,
T.
,
H. E.
Hurlburt
, and
E. J.
Metzger
,
2011
:
Anomalous tropical ocean circulation associated with La Niña Modoki
.
J. Geophys. Res.
,
116
,
C12001
, doi:.
Singh
,
A.
,
T.
Delcroix
, and
S.
Cravatte
,
2011
:
Contrasting the flavors of El Niño and Southern Oscillation using sea surface salinity observations
.
J. Geophys. Res.
,
116
,
C06016
, doi:.
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
, doi:.
Solidoro
,
C.
,
V.
Bandelj
,
P.
Barbieri
,
G.
Cossarini
, and
S.
Fonda Umani
,
2007
:
Understanding dynamic of biogeochemical properties in the northern Adriatic Sea by using self-organizing maps and k-means clustering
.
J. Geophys. Res.
,
112
,
C07S90
, doi:.
Takahashi
,
K.
,
A.
Montecinos
,
K.
Goubanova
, and
B.
Dewitte
,
2011
:
ENSO regimes: Reinterpreting the canonical and Modoki El Niño
.
Geophys. Res. Lett.
,
38
,
L10704
, doi:.
Trenberth
,
K. E.
,
1997
:
The definition of El Niño
.
Bull. Amer. Meteor. Soc.
,
78
,
2771
2777
, doi:.
Wang
,
C.
, and
R. H.
Weisberg
,
1994
:
On the “slow mode” mechanism in ENSO-related coupled ocean–atmosphere models
.
J. Climate
,
7
,
1657
1667
, doi:.
Wang
,
C.
, and
X.
Wang
,
2013
:
Classifying El Niño Modoki I and II by different impacts on rainfall in southern China and typhoon tracks
.
J. Climate
,
26
,
1322
1338
, doi:.
Wang
,
C.
,
C.
Li
,
M.
Mu
, and
W.
Duan
,
2013
:
Seasonal modulations of different impacts of two types of ENSO events on tropical cyclone activity in the western North Pacific
.
Climate Dyn.
,
40
,
2887
2902
, doi:.
Wang
,
X.
, and
C.
Wang
,
2014
:
Different impacts of various El Niño events on the Indian Ocean dipole
.
Climate Dyn.
,
42
,
991
1005
, doi:.
Wang
,
X.
,
W.
Zhou
,
C.
Li
, and
D.
Wang
,
2014
:
Comparison of the impact of two types of El Niño on tropical cyclone genesis over the South China Sea
.
Int. J. Climatol.
,
34
,
2651
2660
, doi:.
Weng
,
H.
,
K.
Ashok
,
S.
Behera
,
S.
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
, doi:.
Wilks
,
D. S.
,
2006
:
On “field significance” and the false discovery rate
.
J. Appl. Meteor. Climatol.
,
45
,
1181
1189
, doi:.
Wu
,
R.
, and
B.
Wang
,
2002
:
A contrast of the East Asian summer monsoon–ENSO relationship between 1962–77 and 1978–93
.
J. Climate
,
15
,
3266
3279
, doi:.
Xiang
,
B.
,
B.
Wang
, and
T.
Li
,
2013
:
A new paradigm for the predominance of standing central Pacific warming after the late 1990s
.
Climate Dyn.
,
41
,
327
340
, doi:.
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:.
Yu
,
J.-Y.
, and
S. T.
Kim
,
2010
:
Identification of Central-Pacific and Eastern-Pacific types of ENSO in CMIP3 models
.
Geophys. Res. Lett.
,
37
, L15705, doi:.
Yu
,
J.-Y.
, and
S. T.
Kim
,
2013
:
Identifying the types of major El Niño events since 1870
.
Int. J. Climatol.
,
33
,
2105
2112
, doi:.
Yu
,
J.-Y.
,
H.-Y.
Kao
,
T.
Lee
, and
S. T.
Kim
,
2011
:
Subsurface ocean temperature indices for Central-Pacific and Eastern-Pacific types of El Niño and La Niña events
.
Theor. Appl. Climatol.
,
103
,
337
344
, doi:.
Yuan
,
Y.
, and
S.
Yang
,
2012
:
Impacts of different types of El Niño on the East Asian climate: Focus on ENSO cycles
.
J. Climate
,
25
,
7702
7722
, doi:.
Yuan
,
Y.
, and
H. M.
Yan
,
2013
:
Different types of La Niña events and different responses of the tropical atmosphere
.
Chin. Sci. Bull.
,
58
,
406
415
, doi:.
Zhang
,
W.
,
L.
Wang
,
B.
Xiang
,
L.
Qi
, and
J.
He
,
2015
:
Impacts of two types of La Niña on the NAO during boreal winter
.
Climate Dyn.
,
44
, 1351–1366, doi:.
Zhou
,
W.
,
C. Y.
Li
, and
X.
Wang
,
2007
:
Possible connection between Pacific oceanic interdecadal pathway and East Asian winter monsoon
.
Geophys. Res. Lett.
,
34
,
L01701
, doi:.