Subsurface Ocean Indices for Central-Pacific and Eastern-Pacific Types of ENSO Jin-Yi Yu 1*, Hsun-Ying Kao 1, and Tong Lee 2 1. Department of Earth System Science, University of California, Irvine, Irvine, CA 2. Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California April 15, 2009 Submitted to Geophysical Research Letters *. Corresponding author address: Dr. Jin-Yi Yu, Department of Earth System Science, University of California, Irvine, CA 92697-3100. E-mail: jyyu@uci.edu 1
ABSTRACT Subsurface ocean indices are developed to identify two distinct types of El Niño Southern Oscillation (ENSO): an eastern-pacific (EP) type that is associated with basinwide thermocline variation and a central-pacific (CP) type that has its subsurface anomalies confined to a shallow layer in the central Pacific. Temperature anomalies in the upper 100 meters are averaged over the eastern (80 W-90 W, 5 S-5 N) and central (160 E-150 W, 5 S-5 N) equatorial Pacific to construct the EP and CP subsurface indices, respectively, for the period of 1958-2001 using an ocean data assimilation product. It is found that the subsurface indices are more independent to each other and show stronger skewness than the SST-based ENSO indices, both of which are desirable properties for separating the EP and CP types of ENSO. Using the subsurface indices, major ENSO events identified by NOAA during the analysis period are classified into the EP or CP types of ENSO. Most of the strong El Niño events are identified as the EP type while most of the strong La Niña events tend to be the CP type. A reversed tendency is found between weak El Niño and La Niña. Between the two largest El Niño events in the past century, the 1997/98 event is identified as a pure EP event while the 1982/83 event is a weak CP event followed by a strong EP event. The different zonal propagations of SST anomalies between these two events can be explained as a result of the different combination of the EP and CP types of ENSO. This study shows that subsurface ocean indices can be effective in monitoring and separating the EP- and CP-ENSO. 2
1. Introduction An increasing number of recent studies have begun to suggest that there may be two distinct types of El Niño Southern Oscillation (ENSO). One of them is close to the canonical ENSO (Rasmusson and Carpenter 1982) that tends to have its sea surface temperature (SST) anomaly center located in the eastern Pacific. The other type of ENSO has its SST anomaly center located more toward the central Pacific and is detached from the South American coast. It was suggested that these two ENSO types have different anomaly patterns, temporal evolution, teleconnections and should not be treated the same (e.g., Larkin and Harrison 2005; Kao and Yu 2009). Kao and Yu (2009) referred to these two types of ENSO as the Eastern-Pacific (EP) ENSO and the Central-Pacific (CP) ENSO, according to the central location of their SST anomalies. Other terms were also used to describe the less-studied CP-ENSO, mostly to its warm phase, including Dateline El Niño (Larkin and Harrison 2005), "El Niño Modoki (Ashok et al. 2007), and Warm pool El Niño (Kug et al. 2009). In the abovementioned recent studies, the CP- and EP-ENSO were separated with different methods but were all related to SST anomalies. For example, Kug et al. (2009) simply compared the relative strength of the Niño-3 and Niño-4 SST indices to determine if an event should be considered a CP type or an EP type. Ashok et al. (2007) emphasized the out-of-phase relation between the SST anomalies in the central equatorial Pacific and those in the eastern and western Pacific to identify their Modoki El Niño events. Kao and Yu (2009) recognized that SST anomalies of both the EP- and CP-ENSO can spread and overlap over the eastern-to-central tropical Pacific and it is difficult to identify these two ENSO types based directly on the SST anomalies. They, therefore, used a combined 3
regression-eof (empirical orthogonal function) method to identify the EP- and CP- ENSO events. They first removed the tropical Pacific SST anomalies that are regressed with the Niño-4 SST index and then applied the EOF to the residual SST anomalies to obtain the leading pattern of the EP-ENSO. The principal component of the leading EOF was then used to identify the EP-ENSO events. Similarly, the pattern of the CP-ENSO was obtained by applying the EOF to the residual SST anomalies after the anomalies regressed with the Niño1+2 SST index were removed. The principal component of that leading EOF was then used to identify the CP-ENSO events. Kao and Yu (2009) showed that these principal components were useful in identifying the EP- and CP-ENSO events. However, the relatively complicate calculations involved in these SST-based indices make them difficult to be directly used for monitoring the EP- and CP-ENSO events. Kao and Yu (2009) found that the subsurface temperature anomalies associated with the EP- and CP-types of ENSO are different from each other in space. This property suggests that ENSO indices based on subsurface ocean information may be useful in identifying and separating the EP- and CP-ENSO. In this study, we develop two subsurface indices to explore this possibility. For the subsurface ocean information, we use the ocean assimilation product from Simple Ocean Data Assimilation (SODA; Carton et al. 2000). The SODA dataset has 40 levels unevenly divided from 5m to 5374m and covers the global oceans from 75.25 S to 89.25 N with a horizontal resolution of 0.5 x0.5. The period of 1958-2001 is analyzed in this study. 4
2. Compare subsurface ENSO indices to SST-based ENSO Indices We first show in Figure 1 the subsurface ocean anomalies associated with the EPand CP-ENSO. The values shown are the lagged correlation coefficients between the principal components of the leading EOFs from Kao and Yu (2009; as in their Figure 3) and the temperature anomalies averaged between 5ºS and 5ºN in the Pacific. For the EP- ENSO (Figure 1a), large and out-of-phase subsurface temperature anomalies appear on both sides of the Pacific and propagate eastward along the thermocline during the evolution of the event. In contrast, the subsurface evolution of the CP-ENSO (Figure 1b) does not have a basin-wide fluctuation pattern and shows little propagation feature. The initiation, development, and termination of the subsurface temperature anomalies all occur in the central Pacific. The anomalies appear first near the surface and then extend downward to a shallow layer of about 100 meters. This depth is well above the local thermocline, which is about 150-200 meters deep (not shown), indicating that thermocline variation is not involved in the evolution of the CP-type of ENSO. It is noticed from Figure 1 that, at the mature phase of the event, the subsurface temperature variation associated with the EP-ENSO has a nodal point located in the central Pacific where the CP-ENSO produces the largest subsurface temperature anomalies. Therefore, the spatial overlapping between the subsurface temperature anomalies of the EP- and CP- ENSO is smaller than that in their SST anomalies. This property suggests that ENSO indices based on subsurface ocean information may be more capable of separating these two types of ENSO than ENSO indices based on SST anomalies. Based on the largest subsurface temperature anomalies found in Figure 1, we 5
choose to average the upper-100m ocean temperature anomalies in the eastern (80 W- 90 W, 5 S-5 N) and central (160 E-150 W, 5 S-5 N) equatorial Pacific as the subsurface indices for the EP- and CP- ENSO, respectively. This depth is deep enough to include the thermocline in the eastern equatorial Pacific but not deep enough to involve the thermocline in the central equatorial Pacific. Therefore, the cross-basin thermocline variations associated with the EP-ENSO is less likely to contaminate the subsurface index for the CP-ENSO. Figure 2 shows the time series of the EP and CP subsurface indices normalized by their own standard deviations for the period of 1958-2001. A 3-month running mean was applied. The skewness coefficients of the time series are 1.68 for the EP subsurface index and 0.63 for the CP subsurface index. A larger value of positive (negative) skewness denotes more extreme warm (cold) events. The skewness coefficients indicate the preference of the EP-type for large El Niño events and the CP-type for large La Niña events. It should be noted that the skewness coefficients for the principal components of the leading EOFs of SST obtained in Kao and Yu (2009) are 1.18 and -0.45 for the EPand CP-ENSO, respectively. Therefore, the spatial asymmetry between strong El Niño and strong La Niña is more obviously revealed by the subsurface indices than by the SST-based indices (i.e. the principal components). Most strong El Niño events defined by the Niño-3.4 SST index were identifiable from the EP subsurface index of Figure 2a, including the 1972/73, 1976/77, 1982/83, 1991/92, and 1997/98 events. In contrast, most large La Niña events defined by the Niño-3.4 SST index were identifiable only by the CP subsurface index shown in Figure 2b, including the 1973/74, 1975/76, 1983/84, 1988/89, and 1998/99 events. The correlation coefficient between the EP and CP subsurface 6
indices is 0.27, which is smaller than the correlation coefficient between the two principal components for SST (which is -0.48). Our analyses indicate that the subsurface indices are more independent to each other and show stronger skewness than the SST-based ENSO indices, both of which are desirable properties for separating the EP and CP types of ENSO. Besides, the method obtaining the subsurface indices for the two types of ENSO is more straightforward than the SST-based indices. Another interesting thing to note is that, Kao and Yu (2009) found that the principal component for the EP-ENSO is dominated by both a 4-year band and a 2-year band while the principal component for the CP-ENSO is dominated by a 2-year band. We find this periodicity contrast is further amplified in the subsurface indices. The EP subsurface index is found to have only one significant power peak near the 4-year band and the CP subsurface index has a significant peak near the 2-year band (not shown). The reason for this further periodicity contrast is not known yet, but the stronger contrast suggests that the subsurface indices are more capable of separating the EP- and CP-ENSO than the SST-based indices. 3. Identify EP- and CP-ENSO events in 1958-2001 We now link the ENSO events identified by Niño-3.4 SST index to the EP and CP subsurface indices. Figure 3 shows the scattering of the normalized monthly values of EP and CP subsurface index from 1958 to 2001. The size of every dot is proportional to the value of Niño-3.4 SST index in the corresponding month, indicating the strength of ENSO. We use this figure to examine how different strengths of ENSO (measured by Niño-3.4 SST index) are projected onto as the EP- or CP-types of ENSO. The dots are colored red for strong El Niño events (with the Niño-3.4 SST index larger than 1ºC) and 7
blue for large La Niña events (with the Niño-3.4 SST index smaller than -1ºC). Here the ±1ºC value is subjectively chosen to highlight the larger events. It is easy to notice from the figure that almost all the large La Niña events are projected onto the coordinate of CP subsurface index; with the subsurface index value changes approximately proportioned to the Niño-3.4 index value. For these events, the EP subsurface index values are in general small. This feature confirms again that strong La Niña events tend to appear as the CPtype. A similar tendency can be found for large El Niño events. A large portion of the red dots is projected onto the coordinate of EP subsurface index with smaller CP subsurface index values. This tendency again shows that most strong El Niño events tend to appear as the EP-type. We next develop criteria for the subsurface indices to identify EP- and CP-ENSO events. During the analysis period of 1958-2001, NOAA has identified 12 El Niño events and 10 La Niña events based on 3-month running mean of Niño 3.4 SST index that is termed Oceanic Niño Index (ONI). These events are listed in Table 1. To be considered as an ENSO event by NOAA, the ONI value has to be larger than 0.5ºC for five consecutive months. We find that if we want to use the subsurface indices to identify all these events, the absolute values of the EP/CP subsurface index have to be larger than 0.5 standard deviation. The standard deviation is 1.3ºC for the EP subsurface index and 0.55ºC for the CP subsurface index. Using this criterion, we are able to classify each of the ONI-identified events into an EP or CP-type of ENSO. The classification is also shown in Table 1. The table shows that, in many events, both the EP- and CP-types of ENSO occurred. For example in 1982/83, both a strong EP-type El Niño and a weak CPtype of El Niño occurred in the tropical Pacific. Upon close inspection (see Figs. 2a and 8
2b), it is noticed that the 1982/83 warming is a weak CP event followed by a strong EP event, rather than just as a single event. In contrast, the strong 1997/98 El Niño is a pure EP event. The 1982/83 event is known to have an eastward propagation of SST anomalies while the propagation for the 1997/98 event tends to be more westward or stationary. We can explain this difference as a result of the different combination of the EP and CP types of ENSO in these two events. For the 1982/83 event, a CP event occurred first and was followed by an EP event. Therefore, the SST anomalies appear spread eastward. For the 1997/98 event, it is a pure EP event that resembles the canonical ENSO event to spread the SST anomalies westward. We further sort the identified ONI events according to their strengths. An event is considered a strong one if the absolute value of its subsurface index is larger than two standard deviations; otherwise the event is considered a weak one. Table 2 summarizes all the strong and weak EP- and CP-ENSO events based on this criterion for the period of 1958-2001. The table shows that all the strong El Niño events are the EP type, while most of the weak El Niño events are the CP type. On the other hand, all strong La Niña events are the CP type while more weak La Niña events tend to be the EP type. The spatial asymmetry between El Niño and La Niña is apparently reversed between strong ENSO events and weak ENSO events. This is consistent with the finding reported recently by Sun and Yu (2009), in which they analyzed the decadal modulation of ENSO SST intensity for the period of 1880-2006 and noted a similar reversal of spatial asymmetry in ENSO SST anomalies. They argued that this reversal is important for giving rise to a 10-15 year modulation cycle of ENSO. 9
Based on events identified in Table 2, we constructed composite lifecycle for the strong EP- and CP-ENSO. The month of peak ENSO intensity of every event is used as the center month of the composite. During the composite analyses, we notice that a strong CP-type La Niña often follows a strong EP-type El Niño. This tendency can be identified from Table 2. For instance, the strong CP-type La Niña events of 1973/74, 1983/84, and 1998/99 came after the strong EP-type El Niño events of 1972/73, 1982/83, and 1997/98, respectively. The composite lifecycle is shown in Figure 4. For a strong EP-type of El Niño (Figure 4a), SST anomalies concentrate in the eastern Pacific. After the warm event (about 6 to 18 months after the peak), it comes negative SST anomalies, which originate from the central Pacific and develop into a CP-type of La Niña. For strong CP-type of La Niña (Figure 4b), the negative SST anomalies start from the central Pacific between 160ºE and 120ºW. The event develops and decays in the central Pacific without an evident propagation. About a year before the La Niña, there is an EP-type of El Niño. The composite shown in Figure 4 further confirms the tendency for strong EP-type El Niño and CP-type La Niña to occur subsequently and alternatively. The right panels in Figure 4 display the normalized values of the EP and CP surface indices and Niño-3.4 SST index during the events. They show that the EP subsurface index captures the evolution of warm events better while the CP subsurface index captures the evolution of cold events better. Niño-3.4 index generally has a value between EP and CP indices. Consistent with the suggestion of Trenberth and Stepaniak (2001), the Niño-3.4 SST index is able to represent the general temperature variations in the equatorial Pacific, but is not effective in separating different flavors of ENSO. 10
4. Discussion We have explored in this study the possibility of using subsurface ocean temperature indices to identify the EP- and CP-types of ENSO. Since the EP-ENSO is associated with basin-wide thermocline variation and the CP-ENSO has its subsurface anomalies confined to a shallow layer of the central Pacific, subsurface ENSO indices based on temperature anomalies in the upper 100-meters of the tropical Pacific were defined to identify these two types of ENSO for the period of 1958-2001. Our analyses focus on linking the ENSO events identified by NOAA's ONI index to the EP and CP types. It should be noted that the subsurface indices also identify some events that the SST index fails to capture. For example, in May 1966 (not shown here), the warm SST shows up in the equatorial Pacific, but was not captured by the Niño-3.4 index. However, the EP subsurface index captures the heat content variations near the coastal region during this event. Also in January 1991, the Niño-3.4 index did not capture the warming in the central Pacific, west to the Niño-3.4 region, but the CP subsurface index is able to reveal this central Pacific event. The criterion used here for the event identification is subjective and not necessary the optimal one. When more understandings are obtained with the linkage between the subsurface evolutions of the EP/CP-ENSO and their underlying physics, further refinements should be made on the subsurface indices to increase their value in monitoring, understanding, and potentially predicting the EP- and CP-types of ENSO. Acknowledgments. The research was support by NASA (NNX06AF49H) and JPL (subcontract No.1290687). Data analyses were performed at University of California, 11
Irvine's Earth System Modeling Facility. 12
References Ashok K., S. Behera, A. S. Rao, H. Weng, T. Yamagata, 2007: El Niño Modoki and its teleconnection. J. Geophys Res., 112, C11007, doi:10.1029/2006jc003798 Carton, J. A., G. A. Chepurin, X. Cao, and B. S. Giese, 2000a: A simple ocean data assimilation analysis of the global upper ocean 1950-1995, Part I: Methodology, J. Phys. Oceanogr., 30, 294-309. Kao, H.-Y., and J.-Y. Yu, 2009: Contrasting Eastern-Pacific and Central-Pacific Types of ENSO. J. Climate, 22, 615 632. 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., 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. Rasmusson, E. M. and T. H. Carpenter, 1982: Variations in tropical sea-surface temperature and surface wind fields associated with the Southern Oscillation El-Niño. Mon. Wea. Rev., 110, 354-384. Sun F, Yu J.-Y., 2008: A 10 15year Modulation Cycle of ENSO Intensity. J. Climate, In 13
Press. Trenberth, K. E., D. P. Stepaniak, 2001: Indices of El Niño evolution. J. Climate, 14, 1697-1701. 14
Table 1: Linking ENSO Events in 1958-2001 to the EP and CP Types El Niño Types La Niña Types (1) 7/63-1/64 CP (1) 4/64-2/65 EP/CP (2) 6/65-4/66 EP/CP (2) 12/67-4/68 EP (3) 11/68-6/69 CP (3) 7/70-1/72 EP/CP (4) 9/69-1/70 CP (4) 5/73-7/74 EP/CP (5) 5/72-3/73 EP/CP (5) 9/74-5/76 EP/CP (6) 9/76-2/77 EP/CP (6) 10/84-9/85 EP/CP (7) 9/77-2/78 CP (7) 5/88-5/89 EP/CP (8) 5/82-6/83 EP/CP (8) 9/95-3/96 EP (9) 9/86-2/88 EP/CP (9) 7/98-6/00 EP/CP (10) 5/91-7/92 EP/CP (10) 10/00-2/01 EP (11) 7//94-3/95 EP/CP (12) 5/97-4/98 EP Table 1. Classification of the NOAA-identified El Niño and La Niña events into the EP or CP types of ENSO according to the subsurface ocean indices. The duration of ENSO events are denoted in month/year. 15
Table 2: Classification between Strong and Weak ENSO EP CP Strong El Niño Weak El Niño Strong La Niña Weak La Niña 72-73, 82-83, 91-92, 97-98 65-66, 76-77, 86-87, 87-88, 94-95 None 64-65, 67-68, 70-71, 71-72, 73-74, 74-75, 75-76, 84-85, 88-89, 95-96, 98-99, 99-00, 00-01 None 63-64,65-66, 68-69, 69-70, 72-73, 76-77, 77-78, 82-83, 86-87, 87-88, 91-92, 94-95 73-74, 75-76, 83-84, 88-89, 98-99 64-65, 70-71, 71-72, 74-75, 75-76, 84-85, 99-00, 00-01 Table 2. Classification of NOAA-defined ENSO events into strong and weak EP/CP events based on the EP and CP subsurface indices during 1958-2001. See text for the definitions of strong and weak events. 16
(a) EP-ENSO (b) CP-ENSO -6-6 -3-3 0 0 3 3 6 6 9 9 12 12 Figure 1. Lagged correlation coefficients between the principal components of the leading EOF for the (a) EP-ENSO (b) CP-ENSO from Kao and Yu (2009) and the subsurface ocean temperature anomalies averaged between 5ºS and 5ºN. Contour intervals are 0.3. The time lag is shown in the lower left corner of each panel. 17
Figure 2. Time series of the normalized (a) EP and (b) CP subsurface index. The dashed lines denote ± 1 standard deviation. 18
Figure 3. Scattering plot of standardized monthly values of the EP and CP subsurface indices from 1958 to 2001. The dot size is proportional to the value of Niño-3.4 SST index in the corresponding month. The coordinate is the value of EP subsurface index. The abscissa is the value of the CP subsurface index. The dot is colored red if its Niño- 3.4 SST index is larger than 1ºC, blue if the SST index is smaller than -1ºC, and green if the SST index is between -1ºC and 1ºC. 19
(a) Composite of strong EP-type of El Niño (b) Composite of strong CP-type of La Niña Figure 4. Composite lifecycle of SST anomalies along the eqiatorial Pacific for (a) strong EP-type of El Niño events and (b) strong CP-type of La Niña events. The center month of the composite is the month of peak value of the subsurface index. Contour intervals are 0.4 C. The right panels show the corresponding EP subsurface index (blue), CP subsurface index (red), and Niño-3.4 SST index (black). 20