We investigated the rapid termination of the 2006 El Niño using observed data and a coupled ocean-atmosphere general circulation model (CGCM). Observed data showed development of an anomalous South Indian Ocean (SIO) anticyclone associated with the concurrent Indian Ocean Dipole during September–November 2006. At the end of 2006, the eastward retreat of the SIO anticyclone in response to the Indian Ocean warming generated an easterly anomaly over the western Pacific, and the resultant oceanic upwelling Kelvin wave terminated the El Niño. CGCM experiments indicated that the formation of the SIO anticyclone was influenced remotely by the Pacific Ocean SST anomalies, and that the easterly anomaly over the western Pacific at the end of 2006 was reinforced by the Indian Ocean SST anomalies. These results suggest that interactive feedback between the Indian and Pacific Oceans helped the rapid termination of the 2006 event.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 El Niño–Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD) [Saji et al., 1999] are the most striking climatic variations in the tropical Indo-Pacific Ocean on an interannual timescale. Since they often occur at the same time, their relationship has been argued. In particular, the role of ENSO with respect to IOD is controversial. Some researchers have pointed out that the development of IOD is strongly controlled by the variations in surface heat and momentum fluxes associated with ENSO [e.g., Lau and Nath, 2003], though other researchers consider that the IOD is an inherent oscillation in the Indian Ocean (IO), and thus independent of ENSO [e.g., Saji and Yamagata, 2003]. In addition, recent studies using observed data and models suggest a possible impact of the IO on ENSO. For example, Kug and Kang  indicated that a positive IOD (pIOD), followed by IO warming, produces an easterly wind stress anomaly over the western edge of the Pacific Ocean (PO) during the mature phase of El Niño, leading to a rapid termination of El Niño. These results suggest that, if the development of IOD is strongly controlled by ENSO, the interaction between the IO and PO can act as a negative feedback mechanism for ENSO. Hence, we need to investigate the interactive processes between the two oceans in order to understand the mechanism for a decaying El Niño in more detail.
 The 2006 El Niño, which was accompanied by a pIOD and terminated rapidly [McPhaden, 2008], provides us with an additional opportunity to understand the interaction between the two oceans. While McPhaden  indicated the role of intraseasonal Madden-Julian Oscillation (MJO) [Madden and Julian, 1972] in the rapid termination of the 2006 El Niño, the present study focuses on the role of the interactive processes between the IO and PO, using observations and a coupled ocean-atmosphere general circulation model (CGCM).
2. Data and Model
 To analyze atmospheric variability, we used monthly mean data from a Japanese 25-year reanalysis [Onogi et al., 2007] and from the Japan Meteorological Agency (JMA) Climate Data Assimilation System (JCDAS). Monthly values of National Oceanic and Atmospheric Administration (NOAA) outgoing longwave radiation (OLR) [Liebmann and Smith, 1996] were used as a proxy for the strength of precipitation or cumulus convection. We also used monthly mean sea surface temperature (SST) data from the JMA operational SST dataset (COBE-SST) [Ishii et al., 2005]. Anomalies were derived after removing climatology for the period from 1979 to 2007. For the composite study, we selected five El Niño events (1982/83, 1986/87, 1991/92, 1997/98, and 2002/03).
 The CGCM employed is the JMA/Meteorological Research Institute (MRI)-CGCM [Yasuda et al., 2007]. The atmospheric component is based on the JMA/MRI atmosphere model [Mizuta et al., 2006]. The model has triangular truncation 95 with a linear Gaussian grid (TL95) in the horizontal, and 40 levels in the vertical direction with the model top at 0.1hPa. The oceanic component is an MRI community ocean model [Ishikawa et al., 2005]. The model domain is from 75°S to 75°N with a horizontal resolution of 1° in longitude and 1° (0.3° near the equator) in latitude. The model has 50 vertical levels (24 levels in the upper 200 m). Adjustments for heat and momentum fluxes between atmospheric and ocean models effectively suppress an initial shock during the early stage of the coupling. Initial atmospheric and oceanic conditions are respectively derived from JCDAS and the Multivariate Ocean Variational Estimation System [Usui et al., 2006]. The skill (anomaly correlation) of the Niño-3.4 (5°S–5°N, 170°W–120°W) SST forecasts over the period 1986–2003 exceeds 0.7 for lead times up to 6 months. Most predictions track the observed anomaly well for the first half year of hindcast. For example, the 1986/87 El Niño simulated by this model was not accompanied with a pIOD, as in the observation [Meyers et al., 2007].
3. Evolution of the 2006 El Niño and IOD
 The 2006 El Niño occurred late, terminated early and was below average strength, compared to past events [McPhaden, 2008, Figure 2]. The SST distribution during September–November 2006 (Figure 1a) indicates relatively large positive anomalies in the central equatorial Pacific along with far-eastern regions, unlike the SST composite map of past events (Figure 1e). Consistent with these SST anomalies, enhanced convective activities were found over the central equatorial Pacific (Figure 1b), whereas they were found widely from the central to the eastern equatorial Pacific in the composite map (Figure 1f). It is noted that convective activities over the Maritime Continent were substantially suppressed in the 2006 event.
 In the Indian Ocean, a pIOD occurred in the boreal summer, peaked in the boreal autumn, and ended in the boreal winter [Horii et al., 2008]. Associated with the development of the pIOD, an easterly wind anomaly developed prior to the enhanced negative SST anomaly in the eastern equatorial Indian Ocean (EEIO) [Vinayachandran et al., 2007]. This enhanced easterly wind anomaly constitutes a part of the South Indian Ocean (SIO) anticyclone [Wang et al., 2003], which occurs as a Rossby wave response to significantly suppressed convective activities over the Maritime Continent (Figure 1b). An SIO anticyclone was also found in the composite map, though the intensity was not as strong as in the 2006 event (Figure 1f). The resultant SST dipole anomaly can in turn enhance the easterly wind anomaly along the equatorial IO through the Bjerknes feedback [Saji et al., 1999].
 At the end of 2006, overall warming occurred in the IO (Figure 1c). This warming can be explained by a relationship between the climatological and anomalous wind [Nagura and Konda, 2007]. Climatological surface zonal winds switch from easterly in boreal autumn to westerly in boreal winter over the EEIO. As a result, the easterly wind anomaly associated with the SIO anticyclone weakened the wind speed, evaporation and coastal upwelling in the EEIO; it warmed the EEIO, and consequently contributed to the overall IO warming. The IO warming induced vigorous convection over the IO (Figure 1d). At the same time, the SIO anticyclone retreated to the east (Figure 1d), presumably due to the atmospheric moist-Kelvin wave response to the convective anomaly in the IO [Annamalai et al., 2005]. The resultant easterly wind anomaly over the western Pacific (WP) along the northern edge of the anticyclone generated an oceanic upwelling Kelvin wave that brought the 2006 El Niño to an end [McPhaden, 2008].
 Meanwhile, the corresponding OLR composite map (Figure 1h) indicates the development of an anomalous Philippine Sea anticyclone, which is also a response to IO warming and could possibly operate as a negative feedback mechanism for ENSO [Weisberg and Wang, 1997; Watanabe and Jin, 2002; Wang et al., 2003]. In the 2006 case, however, the Philippine Sea anticyclone did not develop explosively. Warming east of the Philippines during the 2006/07 winter (Figure 1c), which differs in the composite map (Figure 1g), may have weakened the local air-sea interaction necessary for the maintenance of the anticyclone [Wang et al., 2000].
4. CGCM Experiments in 2006
 In this section, we assess the relative importance of remote (PO) versus local (IO) SST anomalies in forming the SIO anticyclone in 2006 and its impact on the subsequent development of an El Niño, using a CGCM. Forecasting started at an early stage of the event, on 30 July 2006, and was integrated for 12 months. This experiment was referred to as the control experiment (CNTL run). Two additional experiments (“PC run” and “IC run”) were performed to investigate a possible interaction between the PO and IO. In the “PC (IC) run”, climatological SSTs in the PO (IO) were prescribed so as to isolate the role of the IO (PO) from the PO (IO).
Figures 2a and 2b depict simulated SST and OLR anomalies for September–November 2006 from the CNTL run. As in the observed SST anomaly (Figure 1b), the CNTL run captured positive SST anomalies in the equatorial PO, as well as the SST dipole in the IO. Also, the simulated OLR anomalies captured characteristics of the observed ones, namely, enhanced convection over the central equatorial PO and western equatorial Indian Ocean (WEIO) along with suppressed convection associated with the SIO anticyclone.
Figures 2c and 2d illustrate the impact of the PO SST (CNTL run minus PC run) during September–November 2006. The anomalous positive SST in the WEIO and negative SST in the EEIO indicate that the PO SST anomalies promoted the development of pIOD. Anomalous enhanced (suppressed) convective activities over the WEIO (off Sumatra), as well as the easterly wind anomaly associated with the SIO anticyclone over the EEIO, were closely linked to those anomalous SST distributions. This means that the SIO anticyclone was influenced remotely by the PO SST pattern, specifically by positive anomalies localized in the central equatorial PO. On the other hand, the IO SST anomalies (CNTL run minus IC run) did not significantly strengthen suppressed convection over the Maritime Continent (Figures 2e and 2f); the impact of the IO on the SIO anticyclone is, consequently, smaller than that of the PO. However, its intensity was partially reinforced by the IO SST anomalies, suggestive of the Bjerknes feedback.
 The difference in the SIO anticyclone between the CNTL run and IC run caused a difference in the subsequent development of El Niño. To examine it, we used the Niño-3 (5°S–5°N, 150°W–90°W) SST index instead of the Niño-3.4 index, because the model Niño-3 index exhibited the difference between the two runs more clearly. The Niño-3 SST index decreased rapidly into negative values in the CNTL run, as in the observation (Figure 3a). On the other hand, it kept a positive value and gave no indication toward the decay of the El Niño in the IC run. This difference is attributed to an easterly wind anomaly, intensified by the IO SST anomalies, over the WP in the CNTL run (Figure 3b). These results suggest that interaction between the IO and PO through the SIO anticyclone operated as a negative feedback mechanism for the 2006 El Niño.
 One may notice that the predicted changes were not strictly as abrupt as in observations (Figure 3a). This deficiency is probably because a CGCM is missing processes needed to realistically generate a MJO, whose role has been discussed by McPhaden . Short-term upper-ocean variability revealed by recent observations in the equatorial IO [Masumoto et al., 2008] may also imply enhanced MJO activity during 2006. Nevertheless, it is plausible that the IO feedback was at work during the 2006 event. Figure 3b strongly suggests that the IO SST anomalies produced conditions favorable for further enhancement of the MJO-induced easterlies over the WP. Hence, we may say that the IO warming facilitated the rapid termination of the 2006 event.
 We examined the rapid termination of the 2006 El Niño using observed data, and found that the eastward retreat of the SIO anticyclone in response to the IO warming generated an easterly wind anomaly over the WP at the end of 2006. CGCM experiments also indicated that the formation of the SIO anticyclone was influenced remotely by the PO SST anomalies, and that the easterly anomaly over the WP at the end of 2006 was reinforced by the IO SST anomalies, helping the rapid termination of the 2006 event. This implies that the IO feedback played a notable role in the ENSO termination, in addition to the intraseasonal MJO dynamics [McPhaden, 2008], and thus our results strengthen the recent findings [Kug and Kang, 2006; Kug et al., 2006a, 2006b; Dommenget et al., 2006; Ohba and Ueda, 2007].
 The reason for the rapid termination of the 2006 El Niño is presumably linked to localized warming in the central equatorial PO (Figure 1a). This SST pattern, which is in part related to the below average strength of the El Niño, favors enhanced convection over the central equatorial Pacific and suppressed convection over the Maritime Continent. The SIO anticyclone associated with the anomalous convection promotes the development of the pIOD, partially reinforced by the SST dipole anomaly through the Bjerknes feedback during the boreal autumn. The enhanced SIO anticyclone contributes to overall IO warming, leading to an easterly wind anomaly over the WP during the boreal winter. Similar features, including the SST pattern with a concurrent pIOD and its rapid termination of El Niño, are found in the 1994 event, although the cause of the rapid termination needs further investigation.
 The authors thank Ichiro Ishikawa, Ikuo Yoshikawa, Yuhei Takaya, Koichi Ishikawa, and Tsurane Kuragano at Climate Prediction Division, Japan Meteorological Agency for providing observed data and supporting this study. The authors also thank Hiroshi Ishizaki and Tatsuo Motoi for their comments; and Masafumi Kamachi for his careful reading of the manuscript. Comments and suggestions from two reviewers were helpful for the improvement of this paper. This work was funded by the Meteorological Research Institute. Partial support by MEXT Grant-in-Aid for Scientific Research (19540469) was greatly acknowledged.