Transmission of ENSO signal to the Indian Ocean

1. Introduction

[2] The Indo-Pacific atmospheric linkage has been documented by several early studies [e.g., Cadet, 1985; Reason et al., 2000]. In terms of oceanic teleconnection, variability of the Pacific Ocean is known to be transmitted to the Indian Ocean; however, the process throughout the whole Indo-Pacific basin has not been fully explored. The central issues are the manner in which ENSO signals are transmitted, where they come from, and where they go. In a recent study utilizing eXpendable BathyThermograph (XBT) sections since 1983, Wijffels and Meyers [2004] showed that part of the thermal variance on seasonal and interannual time scales in the eastern Indian Ocean can be accounted for by Kelvin and Rossby waves generated by remote zonal winds along the equator of the Pacific Ocean, while much of the rest of the variance is generated by winds over the Indian Ocean. They provided observational evidence that variations in zonal Pacific equatorial winds force a response in the Indian Ocean, primarily along the WA coast [Clarke, 1991; Clarke and Liu, 1993; Meyers, 1996; Masumoto and Meyers, 1998; Potemra, 2001; Potemra et al., 2002]. These previous studies suggest that the energy off WA arises from equatorial Rossby waves generated by zonal wind anomalies in the central equatorial Pacific which become coastally trapped waves where the New Guinea coast intersects the Pacific equator. Using an XBT section from Fremantle to Java, Cai et al. [2005] examined the relationship between temperature on the section and variability of the equatorial Pacific, in the context of the ENSO recharge-oscillator paradigm [Jin, 1997; Meinen and McPhaden, 2000] and confirmed the transmission to the central WA coast. Because the XBT data is limited to a line, a detailed pathway is not fully established. In this study, we examine the pathway further using a thermal analysis covering the whole Indo-Pacific domain during 20 years since 1980.

2. Data and Method

[3] The thermal analysis is the operational subsurface temperature analysis from the Australian Bureau of Meteorology Research Center [Meyers et al., 1991; Smith, 1995a, 1995b], which is an optimal interpolation on a 1° latitude by 2° longitude grid at 14 depth levels, throughout the Indo-Pacific basin. It is based primarily on XBT profiles and time-series from the Tropical Atmosphere Ocean (TAO) buoy array [McPhaden et al., 1998], and covers a 20-year period since 1980. The data in the Pacific Ocean domain has been used in numerous studies [Meinen and McPhaden, 2000; Cai et al., 2004; Kessler, 2002]. The relationship of oceanic variability to the wind field is documented with data from the NCEP reanalysis [Kalnay et al., 1996]. To focus on interannual time scales, a 13-month running mean operation is applied before analysis.

[4] We apply a complex empirical orthogonal function (CEOF) analysis [Cai and Baines, 2001], which depicts the equatorial Pacific discharge/recharge process and the propagation of off-equatorial Rossby waves, to the filtered thermocline anomalies in the whole Indo-Pacific domain. Since variance of thermocline anomalies is much greater in the equatorial than in the off-equatorial region, to enhance the spatial coherence, the grid point anomalies are normalized to have unit variance. The CEOF analysis brings out additional features at higher latitudes associated with the ENSO processes. The spatial pattern is further documented using a simple lag-correlation analysis.

3. ENSO Discharge/Recharge Signals Within the Indo-Pacific System

[5] According to the recharge-oscillator paradigm [Jin, 1997], El Niño is preceded by a build up of heat content with a deepened thermocline, and followed by a heat deficit with an anomalously shallow thermocline, which in turn precedes a La Nina event. These properties were validated by observations of the upper Pacific Ocean heat content over the past two decades by Meinen and McPhaden [2000]. They showed that the first and second EOF of equatorial Pacific thermocline anomalies (their Figure 3), represented by anomalies of 20°C isotherm depth (D20), account for a similar percentage of the total variance, and together EOFs 1 and 2 describe ENSO cycles with a phase lag of approximately 90°. EOF1 has a zonal dipole structure representing the mature phase of ENSO, and EOF2 a zonal symmetric pattern indicating the recharge/discharge phase. The CEOF analysis discussed below allows an exploration of the transition between these two EOFs. Although we focus on El Niño and the associated discharge, the features discussed below apply to the opposite phase: La Niña and the associated recharge.

[6] Figure 1 displays half a cycle of the leading CEOF mode of the D20 anomalies, which accounts for 53% of the total variance, at a phase interval of 22.5° corresponding to a time interval of approximately three months for a four-year ENSO cycle. At phase 0°, the pattern in the equatorial Pacific (10°S–10°N) corresponds to the recharged phase, and at phase 90°, the pattern corresponds to the mature phase of an El Niño with a zonal dipole in the equatorial Pacific, and at phase 180°, to the discharged phase.

[7] The extension to the off-equatorial region brings out several additional features. At phase 0°, upwelling Rossby wave (indicated by negative contours) develops and radiates from the eastern boundary. At phase 22.5°, the upwelling Rossby wave is reinforced in the off-equatorial NP in the vicinity of (155°W, 17°N). Subsequently, the Rossby wave experiences a strong growth south of 10°N, and propagates westward. It impinges on the western boundary, moves equatorward and then reflects back on the equator as an equatorial Kelvin wave (phase 25° to phase 67.5°). This equatorward movement was first identified by Godfrey [1975], who named it the Kelvin-Munk wave, and was confirmed by the McCreary [1983] study. The reflected Kelvin wave then switches into a coastally-trapped wave and propagates poleward along the WA coast, a path discussed by previous studies [Clarke, 1991; Clarke and Liu, 1993; Meyers, 1996; Masumoto and Meyers, 1998; Potemra, 2001; Potemra et al., 2002], contributing to a discharged phase of an El Niño at phase 180°. The evolution described above supports that the recharge-oscillator paradigm is consistent with the delayed-oscillator paradigm [Jin, 1997].

[8] Upon reaching the central WA coast, the discharge radiates westward into the Indian Ocean 20°–30° of longitude. Thus an increasingly large area of the Indian Ocean is seen to participate in the discharge process. The discharge off WA reaches a maximum approximately three months after an El Niño peaks (phase 112.5°). Thus some of the signal along the central WA coast can be traced to the off-equatorial NP in the vicinity of (155°W, 17°N), where the Rossby wave initially appeared. This NP transmission “ray-path” is further discussed below in a simple correlation analysis.

4. The Off-Equatorial NP Ray-Path

[9] The off-equatorial NP ray-path is further documented by a lag-correlation analysis of D20 and zonal wind stress anomalies with time series of D20 anomalies averaged over a central WA box (Figure 2). To compare with Figure 1, the time series is sign-reversed so that a discharge signal is represented by negative correlation. Maximum discharge on the WA coast appears at Lag 0, and at Lag -3 it corresponds to an El Niño peak. Overall, there is strong similarity between Figures 1 and 2, suggesting that the WA anomaly is predominantly generated by ENSO processes.

[10] Some 15 months prior to the maximum of discharge along the WA coast (Lag -15), the D20 anomaly pattern resembles that of Figure 1a, and the equatorial Pacific is at a recharged phase. To the east of Hawaii (20°N), negative D20 anomalies develop with reinforcing westerly anomalies immediately south. The anomalies strengthen and propagate westward for the next 6 months (Lag -12 to Lag -9) accompanied by the westerly anomalies. In the mean time, westerly anomalies develop over the Indonesian region (Lag -15) and subsequently in the equatorial Western Pacific (Lag -9) with a maximum at about 5°N (asymmetric about the equator) and join the westerly patch south of Hawaii. The westerly anomalies then move eastward and southward, reinforcing the discharge, especially between 10°N and the equator. There is no southern hemispheric counterpart.

[11] At Lag -3, the pattern is similar to that of the mature phase of El Niño (Figure 1e). By this time, the intense discharge near the western boundary reaches the northwest coast of Australia. The westerly anomalies over the equatorial Pacific have become more or less symmetric about the equator, and a response in the thermocline begins to develop in the Southern Hemisphere. The equatorial westerlies generate a Rossby wave in the southern off-equatorial western Pacific (SEWP) region around 160°E. The evolution indicates that in some El Niño events easterly anomalies develop over the equatorial eastern Indian Ocean and sometimes contributes to the development of an Indian Ocean Dipole (IOD) event [Yamagata et al., 2004, Feng et al., 2001]. The easterly anomalies raise the thermocline along the Sumatra-Java coast [Wijffels and Meyers, 2004], and in the months from June to November induce an enhanced evaporative heat loss [Hendon, 2003]. Both processes are conducive to the development of cold anomalies off the Sumatra-Java coast. Note that NOT all El Niño events are associated with an IOD; this is underscored by the small anomaly off the Sumatra-Java coast.

[12] At Lag 0, approximately 3 months after the mature phase of El Niño, the signal along the central WA coast reaches a maximum. The westerly anomalies in the equatorial Pacific have moved further south [Harrison and Vecchi, 1999], the maximum residing at the southern equatorial latitudes. Meanwhile, easterlies develop in the eastern Indian Ocean and the far western equatorial Pacific [Wang, 2001]. Equatorial Kelvin waves generated by these processes and by the off-equatorial Rossby wave reflection contribute to the demise of the warm condition in the eastern Pacific.

[13] One of the surprising results of this analysis is that the discharge in the Indian Ocean appears to be linked to the region near Hawaii (Figures 2d and 2e). The link is further illustrated in Figure 3, which plots the time series of D20 anomalies along the central WA and averaged over an off-equatorial NP box (17°N–22°N, 155°W–150°W). Because the NP time series leads the WA time series by about 9 months, the former is shifted forward by 9 months. A strong coherence is seen (with a correlation of 0.75 at a 9-month lag, meaning that at least 55% of the interannual variance of the WA thermocline anomaly is linked to that near Hawaii). The coherence is further highlighted by a common lack of strong D20 anomalies during the 1997/1998 El Niño. A strong discharge in the 1997/98 event was confined in the equatorial Pacific [Meinen and McPhaden, 2000, Figure 3].

5. Comparison With Previous Studies

[14] Previous studies [e.g., Wijffels and Meyers, 2004] find that there is little lag between anomalies along the central WA coast and the equatorial Pacific wind. Figure 4 plots the central WA D20 time series (thick solid curve) and zonal wind anomalies averaged over Niño3.4 region (thin solid curve). The Niño3.4 (or Niño4) wind leads the WA D20 by one month with a correlation of 0.68. The wind leads D20 averaged over a SEWP (0°–5°S, 155°E–160°E) box (dashed curve, Figure 4) by two months with a correlation of 0.85. (Note that the SEWP time series shows a large recharge/discharge associated with the 1997/1998 event.) The small lag of the WA and SEWP thermocline variations relative to the equatorial Pacific wind is consistent with the Wijffels and Meyers [2004] result, and the small lag among these indices suggests that the NP signal leads all these indices by a similar amount of time, about 7–9 months.

[15] The picture that emerges from Figure 2 is that thermocline anomalies initially develop at NP some 7 months prior to the mature phase of ENSO. As ENSO develops, the off-equatorial zonal wind anomalies join the zonal wind anomalies over the northern western Pacific and move eastward and southward, reinforcing the NP D20 anomalies as they propagate westward and equatorward. Consequently, the peaking of the equatorial Pacific wind and the western boundary reflection occur nearly simultaneously. In the subsequent 2 months, the NP anomaly reaches the WA coast, hence the 9-month lag.

6. Conclusions

[16] The issue of whether the NP is involved in the transmission of ENSO signals to the Indian Ocean was not addressed by previous studies. Our analysis confirms that ENSO discharge/recharge arrives mainly at the central WA coast and finds that the transmission goes through an off-equatorial NP ray-path whereby the anomaly, after an initial development near Hawaii, propagates westward as a Rossby wave, and upon impinging on the western boundary, it moves equatorward along the Kelvin-Munk ray-path proposed by Godfrey, and reflects as an equatorial Kelvin wave. As the reflected Kelvin wave propagates eastward, it impinges on the Australasian continent and moves poleward along the WA coast as a coastally-trapped wave, radiating Rossby waves into the south Indian Ocean. Along the above path, the north Pacific anomaly reaches the central WA coast some 9 months later. On-route to the WA coast the anomaly is reinforced by evolving winds associated with ENSO. Our study also finds that off-equatorial Rossby waves played a small part in the recharge/discharge associated with the 1997/1998 event, with a correspondingly small signal along the WA coast despite the presence of a strong discharge/recharge signal in the equatorial Pacific. We note however that equatorial Kelvin waves generated by easterly anomalies that develop at the mature El Niño phase in the far western equatorial Pacific [Wang, 2001] and by westerly anomalies that move southward from the equator [Harrison and Vecchi, 1999] also contribute to the discharge. An examination of the cause of the weak WA-Hawaii link for the 1997/1998 El Niño is beyond the scope of the present study, and awaits a determination of the relative importance of various ENSO oscillation paradigms in the generation and demise of this event.