El Niño-Southern Oscillation (ENSO), which dominates variability on interannual timescale in the climate system, is known to exhibit various spatio-temporal characteristics. Recent studies show that in additional to a canonical El Niño with its major center of sea surface temperatures (SST) anomalies in the equatorial Pacific cold-tongue region, a different type of El Niño with its major action center shifted to the warm-pool edge has emerged and become more common during the past two decades. Because the SST patterns of these two types of El Niño events are highly correlated, neither of the traditional Niño3 and Niño4 SST indices alone is effective in representing the new-type El Niño. Through a simple transformation of the Niño3 and Niño4 indices, we devised two new indices that separately identify the two types of ENSO events. Unlike the Niño3 and Niño4 indices, the two new indices are of little simultaneous correlation. The SST patterns associated with these new indices capture SST characteristics of the two types of ENSO. Their running lagged-correlations capture different ENSO-phase propagations and ENSO regime changes associated with the climate shift in 1976/77.
 The ENSO phenomenon has been traditionally quantified in terms of simple indices such as SST anomalies in the Niño3 region (5°N–5°S, 150°–90°W) and Niño4 region (5°N–5°S, 160°E–150°W) or a mixture of these two, the Niño3.4 region (5°N–5°S, 170°–120°W). However, the fact that ENSO has different SST patterns makes the Niño3.4 index inadequate to serve as “one size fits all” index to capture these different types of ENSO. Moreover, since the Niño3 and Niño4 indices are highly correlated (e.g., Figure 1a), different new indices were introduced to depict different flavors of ENSO. For example, Trenberth and Stepaniak  first proposed to use the Niño3.4 index as representing the anomalous SST averaged across the equatorial Pacific and a so-called Trans-Niño index (TNI) as representing the difference between the normalized SST anomalies averaged in the Niño1+2 and Niño4 regions. Because the Niño3.4 index and TNI are nearly uncorrelated at zero lag, these two indices were shown to be capable of describing the evolutions and different flavors of ENSO and dramatic ENSO regime changes.
 In an attempt to characterize the new-type ENSO with a single index, Ashok et al.  proposed an El Niño Modoki index (EMI) which may be viewed as a variation of the TNI index. Although different from TNI, EMI actually captures, to a large extent, the second EOF pattern of the equatorial Pacific SST anomalies. Realizing that the second EOF is not quite adequate to describe main features of the new type of ENSO, Kao and Yu  proposed another somewhat complicated index to characterize the new type of ENSO, which is also a EOF-based index but obtained by first linearly removing the SST anomalies related to the Niño1+2 index. All of these somewhat related indices utilized SST anomalies beyond the Niño3 and Niño4 regions despite the fact that these two regions cover most of the equatorial Pacific.
 It is conceivable that the Niño3 and Niño4 indices, which capture the broad-scale nature of the ENSO-associated SST anomalies, contain the main signal of these two different types of ENSO. In fact, Kug et al.  proposed to classify the two types of El Niño events by comparing relative magnitudes of these two indices. Motivated by this method of classification, we introduce two new Niño indices utilizing only the Niño3 and Niño4 indices and a simple nonlinear transformation. We find that the main patterns of the two types of ENSO, different ENSO phase propagations, and ENSO regime changes, can be all conveniently delineated using these two new indices. We thus propose that these two new Niño indices may be used to characterize the large-scale natures of the different flavors of ENSO.
2. Two New Niño Indices
 The Niño3.4 index and TNI (or EMI) are closely associated with the first and second EOF modes of the equatorial Pacific SST anomalies, respectively, and thus these two indices are nearly orthogonal. However, the SST patterns of the warm-pool and cold-tongue El Niños are far from orthogonal and are very similar [Kao and Yu, 2009; Kug et al., 2009; Yeh et al., 2009]. Thus, these indices are not optimal to differentiate the warm-pool (WP) and cold-tongue (CT) types of ENSO. Recently, Kug et al.  and Yeh et al.  classified the two types of El Niño events only using the Niño3 and Niño4 indices. They first identify the years during which the boreal-winter Niño3 and Niño4 indices are above 0.5°C. Among these years, a CT (WP) El Niño year is identified when the winter Niño3 index is greater (smaller) than the winter Niño4 index. They then derived composite patterns of atmospheric and oceanic variables for the two types of El Niño using this method of classification.
 Using anomalous Niño3 and Niño4 SST indices (1950/01 to 2010/06) directly obtained from the website (http://www.cpc.noaa.gov/data/indices/sstoi.indices) of the Climate Prediction Center in National Oceanic and Atmospheric Administration (NOAA), we examine the relationship between the two indices. As seen in Figure 1a, in the phase space of the Niño3 and Niño4 indices, the two indices are strongly correlated with a simultaneous correlation over 0.7. Nevertheless, the two types of El Niño are well separated along the indicated diagonal when both indices have amplitudes greater than 0.5°C. Albeit, the separation of the cold states of the two types of ENSO is much less clear, as noted by Kug et al. .
 Motivated by the clustered features seen in Figure 1a, using a coordinate transform in the N3-N4 phase space, we introduce two new Niño indices: NCT and NWP which are defined as follows:
Here, N3 and N4 denote Niño3 and Niño4 indices, respectively. These two new indices are simply a piecewise linear combination of the Niño3 and Niño4 indices conditioned by the ENSO phase. The parameter of this transformation is determined by a minimization procedure to make the apparent clusters along the indicated diagonal in Figure 1a centered on the new coordinate axes as much as possible, as shown in Figure 1b. Namely, the centers of red- and green-dot clusters are moved away from the diagonal and relocated almost on the positive horizontal and vertical axes. This transformation methodology may be applied to model simulations, although the value of the parameter is data-dependent. In Figure 1b, there are two less pronounced centers of blue-dot cluster as well. But, the separation of the two blue clusters is less evident, which is consistent with the notion that the SST patterns of La Niña events corresponding to the WP and CT ENSOs are not as distinct as their El Niño patterns [Kug et al., 2009]. The separation of two corresponding types of La Niña events, as advocated by Kao and Yu  and Ashok and Yamagata , may not be necessary because the SST patterns during La Niña phases of ENSO cycles generally tend to propagate westward.
 The correlation between NCT and NWP indices is about 0.13 in Figure 1b, which is substantially reduced from a relatively high correlation of 0.73 between the Niño3 and Niño4 indices in Figure 1a. Moreover, the two indices tend to characterize the WP and CT ENSO events, respectively. For example, when the WP (CT) index is positive greater than a threshold value, such as one standard deviation, then the warm SST anomaly may be regarded as a WP (CT) El Niño state. Therefore, we will refer these two new indices (NWP and NCT) as the WP and CT ENSO indices, respectively.
 It should be pointed out that these two new indices are highly correlated with N3 and N4 indices as seen in Figures 2a and 2b. In particular, the difference between N3 and NCT indices are mostly very minor, indicating that N3 index can still capture the CT ENSO well. When the WP El Niño occurs, NCT index becomes smaller than 0.5°C for most cases even though N3 index may still be larger than 0.5°C, such as 1994/95, 2009/2010 events. Here, NCT index tends to be smaller than N3 index so that these events are not classified as CT El Niño events even though they would be if one used only N3 index. The difference between N4 and NWP indices are substantial during the strongest ENSO events, such as 1972/73, 1982/83, 1997/98, despite the fact that N4 and NWP indices are also highly correlated. The basin-wide nature of SST anomaly patterns during these strong events give rise to large amplitudes in both N3 and N4 indices, but not in NWP index. This is consistent with the notion that these events are identified as CT El Niño events. Also, as seen in Figure 2, the intensity of WP (CT) El Niño tends to become stronger (weaker) in the past three decades, as noted by Lee and McPhaden .
 To further illustrate that NCT and NWP indices are indicators for the CT and WP ENSO events, we examine the spatial patterns of the two types of ENSO. We first regress monthly-mean SST anomaly field onto NWP, NCT, N4 and N3 indices, respectively (see Figures 3a–3c). The dataset used is the SST version 3b (ERSST V3b) [Smith et al., 2008] from National Climate Data Center/NOAA. We have omitted the regressed pattern of SST with N3 because it is almost identical to the pattern shown in Figure 3b. We also plotted three composites (Figures 3d–3f) by averaging the SST anomalies corresponding to the state represented by the red, green, and blue points in Figure 1 as the patterns for the CT, and WP El Niño events and all La Niña events, respectively. The WP El Niño patterns captured by either composite or regression with NWP index are very similar to each other. Both differ significantly from the regressed pattern with N4 index because the later is a mixture of the WP and CT events so that N4 index captures the strong CT events as well. One may distinguish the La Niña patterns of the WP and CT events by composing the SST patterns in terms of the clusters centered over the negative sides of axes in Figure 1b. These patterns will be quite similar to the regressed patterns shown in Figures 3a and 3b with reverse signs. Averaging these two patterns will lead to a pattern similar to the composite of all La Niña (Figure 3f). Overall, the two new indices are capable of distinguishing the SST patterns for the two types of ENSO events.
 In comparison with the patterns of the two types of ENSO in previous works, our WP El Niño pattern in Figure 3a (or Figure 3c) is similar to those of Kao and Yu  and Kug et al. . It differs from the pattern of Ashok et al. , which showed a negative-anomaly area located over the southeast equatorial Pacific near the coast of South America. This is because EMI tends to capture the second EOF pattern of SST anomalies, whereas NWP index is derived from the basin-wide SST anomalies captured in the Niño3 and Niño4 regions. Our NWP index is significantly correlated with EMI (0.78) as well, because they both have the signal of zonal contrast of SST anomalies across the equatorial Pacific. However, our NWP index focuses more on the WP SST anomalies near the dateline.
3. ENSO Regimes Delineated by the NCT and NWP Indices
Trenberth and Stepaniak  used Niño3.4 index and TNI to describe the evolutions and different flavors of ENSO and dramatic ENSO regime changes, while Ashok et al.  proposed EMI to describe so-called “El Niño Modoki”. Both TNI and EMI involve SST anomalies over the far eastern tropical Pacific in order to derive more signal of ENSO other than that contained in the Niño3 and Niño4 indices. In this section, we try to demonstrate that NCT and NWP indices, which are derived from Niño3 and Niño4 indices, effectively depict the different flavors of ENSO and the dramatic ENSO regime changes.
 To elucidate this point, we follow the approach of Trenberth and Stepaniak  by examining lag correlation between NCT and NWP indices (see Figure 4). What has been captured by the lag correlation between Niño3.4 index and TNI, as discussed by Trenberth and Stepaniak , is also captured in Figure 4a. For example, the lag correlation, as delineated in their work, revealed the nature of different phase propagation of ENSO events. TNI is clearly leading Niño3.4 index before the noted climate regime shift at 1976/77, which reflects the fact that ENSO has marked westward propagation before that climate regime shift. After the climate shift, ENSO events become either slightly eastward propagating or stationary so that the lead-lag relationship between Niño3.4 index and TNI tends to reverse. Sharp transition between these two different ENSO regimes was clearly seen even though the lag correlations are calculated using a 20-year running window.
 These two features are captured in Figure 4 as well. Initially, NCT index clearly leads NWP index by about 4–5 months with a significant correlation before the climate shift in 1976/77, as shown in Figure 4a. Thus, this cross correlation captures the well-known westward propagation of ENSO SST anomalies before that climate regime shift. However, this strong lag correlation disappears sharply near the time of the climate shift. Some weak positive correlations with NWP index leading NCT index appear with some weak eastward propagation after the regime transition. This suggests that NCT and NWP indices effectively capture the sharp transition between the ENSO regimes in terms of the different zonal propagations in equatorial SST anomalies. Moreover, we calculated the partial lead-lag correlations for positive and negative phases of the two new indices separately. As seen in Figure 4b, the rapid ENSO regime transition from westward to eastward propagation for the warm phase of ENSO occurred at the so-called climate regime shift. At the same time, the westward phase propagation of ENSO cold phase continued after the climate shift, consistent with the result of McPhaden and Zhang .
 Moreover, the results in Figures 4a and 4b reveal another feature that is not evident in the lead-lag correlation between Niño3.4 index and TNI as analyzed by Trenberth and Stepaniak . That is, NCT and NWP indices become nearly independent after the 1976/77 climate regime shift at almost all lags. Although the data is too short to be statistically significant, our result is in support of potential independence of the two types of ENSO. In other words, after the climate regime shift, the new-type WP ENSO emerges in addition to the conventional CT ENSO and either may occur, likely depending on sources of triggering. Thus, our results support the conclusion: The WP ENSO events can be regarded as a new type of ENSO, and it is nearly independent from the CT ENSO. It occurs more frequently in the past two decades as noted in previous works.
4. Summary and Conclusions
 Increasing evidence supports the notion of existence of two prominent types of El Niño: The canonical CT El Niño with maximum surface warming centered in the eastern Pacific, and the WP El Niño that has maximum warming in the central Pacific near the warm pool edge. To better depict the WP ENSO with a simple index, we proposed a WP index for the WP ENSO and also a CT index for the CT ENSO, based on a simple transformation of the traditional Niño3 and Niño4 indices.
 Using the two new indices, we examined the related SST patterns and showed that these patterns capture the SST characteristics of the WP and CT El Niños. The WP index, although related to other indices, such as EMI, does not only captures the WP-ENSO SST pattern, but also capture its relationship with the CT index in describing the ENSO regime change that occurred in 1976/77. In particular, the WP and CT indices are not only effective to characterize the marked changes in terms of zonal propagation of ENSO SST anomalies, but also capture emergence of the new ENSO regime with coexistence of the two independent types of ENSO.
 The possible coexistence of two ENSO-like modes bearing resemblances to the two types of ENSO was noted by Bejarano and Jin . This ENSO regime with two leading modes occurs in the neighborhood of codimension-2 degeneracy that is a ubiquitous property of ENSO regime diagrams in 2-parameter space of theoretical ENSO models [Jin and Neelin, 1993; Jin, 1997]. Whether the WP ENSO events will become more frequent and the regime with the coexistence of the two-types of ENSO will become a prevailing ENSO regime in the future are subjects worthy of much further research. The new WP index proposed here provides a useful addition to the diagnostic and monitoring tools for future studies of the WP ENSO, including its nature and possible changes as well as its distinct impacts, some of which have already been noted [e.g., Weng et al., 2009; Kim et al., 2009; Yu and Kim, 2010].
 We thank two reviewers for their valuable comments and A. F. Z. Levine for proof-reading. This work is jointly supported by National Science Foundation (NSF) grants ATM 1034798, NOAA grant NA10OAR4310200, DOE grant DE-SC0005110, and NSF of China (NSFC) grants 40705021 and 40805028, the National Science and Technology Support Program of China (2007BAC29B03), and the 973 Program of China (2010CB950404).