Patterns of Coherent Climate Signals in the Indian Ocean during the 20th Century

1. Introduction

[2] Global climate signals ranging from 2- to 20-year periods have been identified by Mann and Park [1996], Tourre and White [1995], Allan [2000], and White and Tourre [2003]. These include the quasi-biennial oscillation (QBO), the El Niño-Southern Oscillation (ENSO), the quasi-decadal oscillation (QDO), and the bi-decadal oscillation (BDO). Evidence for these signals have already been observed in the Indian Ocean [Reason et al., 2000]. Interannual patterns of SST and SLP variability are referred to as the ENSO signal [e.g., Tourre and White, 1997; Allan et al., 2001] or the Indian Ocean Dipole (IOD) [e.g., Saji et al., 1999]. One question is whether these signals represent IOD phenomena, particular to the Indian Ocean, or are regional aspects of global QBO, ENSO, and BDO signals?

[3] To answer this question we conduct a joint frequency analysis for SST and SLP anomalies for the Indian Ocean (20°N to 30°S), finding QBO, ENSO, and BDO signals therein over most of the 20th Century. These signals share similar patterns and evolution, as do similar signals over the Pacific and global oceans [Tourre et al., 2001; White and Tourre, 2003]. Comparing results from this paper with recent modelling results [e.g., Baquero-Bernal et al., 2002], we achieve a better understanding of the physical mechanisms associated with these Indian Ocean climate fluctuations. QBO, ENSO and QDO signals in the Indian Ocean are maintained by dynamical mechanisms similar as those governing these signals in the Pacific Ocean [White et al., 2003]. This completes a series of studies by the authors, with a Multi-Taper Method/Singular Value Decomposition (MTM-SVD) analyses applied to SST and SLP anomalies for the Atlantic, Pacific, Indian, and Global oceans [Tourre et al., 1998, 2001; White and Tourre, 2003].

[4] Eastward phase propagation of SST and SLP anomalies across the Indian Ocean in these signals may improve climate prediction over adjacent land surfaces [Joseph et al., 1994; Jury et al., 1996]. Incorporating this information into decision-making schemes, will mitigate regional impacts of recurring drought/flood in east Africa, India, Australia and southeast Asia [Hansen et al., 2001].

2. Data and Results

[5] The MTM-SVD methodology [Mann and Park, 1999] is applied to 92 years (1900–1991) of SST and SLP gridded anomalies in the Indian Ocean [Kaplan et al., 1998]. Standardized anomalies are used, with 3 tapers for reasonable frequency resolution whilst providing sufficient degrees of freedom for signal/noise decomposition. The joint local fractional variance (LFV) spectrum represents fraction of variance in narrow frequency bands. Stability of the analysis was tested by comparing results based upon the full period, with those based on shorter periods (e.g., 1950–1991). Significant frequency bands represent independent information about spatially correlated oscillatory signals; i.e., BDO, ENSO, and QBO [White and Tourre, 2003]. In Figure 1 (top), the LFV spectrum of the first joint SST-SLP singular values is displayed. Significance of peaks in the LFV spectrum are determined through bootstrap temporal re-sampling methods [Efron, 1990]. Potential bias from serial correlation was minimized by permuting time-series 1000 times, randomly and separately, for the 92-yr time sequences for each of the 12 months of the year. We find the BDO signal of 14 to 22-year periods significant at the 95% level, with peak LFV (∼0.75) occurring near 16.7-year period (top, Figure 1). We find three signals significant at the 95% confidence level within the ∼3 to 7-year period ENSO band, with maximum LFV (∼0.87) occurring near 3.4-year period, 4.4-year and 6.2-year period. We also find a significant QBO signal, with peak LFV (∼0.82) occurring near 2.2-year period. The patterns and evolution of each of these signals in SST and SLP variability are displayed in Figures 2 and 3, respectively. Since the 3.4- and 4.4-year period ENSO signals display similar patterns and evolution, we display SST and SLP variability only for the 3.4-year period and 6.2-year period ENSO signals.

[6] To represent the intensity of these signals as a function of time, we display time sequences of associated standardized SST anomalies for the BDO signal (14.2- to 22.2-year periods), the QBO signal (2.1- to 2.4-year periods), and ENSO signals (4.8- to 6.5-year periods; 3.8- to 4.8-year periods; 3.1- to 3.7-year periods) (Figure 1, bottom, 1 to 5). The values are band-pass filtered for the corresponding frequency bands [Allan, 2000]. The time sequence of each narrow-band signal (1 to 5) all display peaks (troughs) corresponding to the central-eastern Indian Ocean warm (cool) phase (Figure 2), modulated for frequency amplitude, and phase. We find these time sequences (Figure 1, 1 to 5) displaying comparable amplitudes (e.g., ∼0.1°C for the QBO signal and ∼0.2°C for the BDO signal). We also find the 3.4-year period ENSO signal diminished in intensity during the 1920s and 1950s, whilst the 2.2-year period QBO signal was diminished in the 1910s, 1940s, and the 1950s for reasons yet to be determined.

[7] To verify QBO and ENSO time sequences (Figure 1, bottom, 2 to 5) we compare their sum to a band-passed (2- to 7-year) Indian Ocean SST index in the central-eastern equatorial region (80°E–95°E; 5°N–5°S). Both indices (i.e., the sum of signals and the SST index) cross-correlate at 0.73, significant at the 99% confidence level. Thus the four signals explain ∼50% of QBO and ENSO variance in the Indian Ocean SST index for the 92-year record.

[8] Evolution of SST patterns of variability for the 16.7, 6.2-, 3.4-, and 2.2-year period signals (Figure 2) are reconstructed for half-cycles (0° to 180° phase), with phases adjusted so that each signal achieves maximum warming in the central-eastern Indian Ocean (180° phase). The 3.4-year period and 6.2-year period ENSO signals are characterized by tropical warming near the Horn of Africa (60° phase), expanding eastward towards the eastern Indian Ocean, and the South China Sea (180° phase). For the QBO and BDO signals, warming first appears in the vicinity of Madagascar (60° phase), propagating northward and eastward towards India and then into the eastern Indian Ocean and South China Sea (150° phase). For the 3.4- and 4.4-year period ENSO signals, the SST evolution is similar to that featured in Tourre and White [1997] and White and Tourre [2003]; i.e., eastward propagation with maxima along the equator at phase speed of 0.20 m s⁻¹. For the 16.7-year period BDO signal and 2.2-year period QBO signal, the SST evolution is similar to that featured in White and Tourre [2003]; i.e., positive SST loadings appear first in the southwest, and then propagate into the eastern tropical Indian Ocean (180° phase).

[9] We find the spatio-temporal evolution of SLP patterns of variability for 16.7-, 6.2-, 3.4-, and 2.2-year period signals (Figure 3) reconstructed for one-half cycle (from 0° to 180° phase) in phase with the evolution of SST patterns of variability (Figure 2), allowing SST and SLP evolution to be compared more easily. The northward and eastward evolution of the 4 signals in SLP variability is striking, with positive SLP weights forming in the southwest Indian Ocean at 60° phase, propagating northward and eastward into the eastern Indian Ocean/Indonesia-Borneo at 150° phase (Figure 3). This corroborates results by Tourre and White [1997] and White and Tourre [2003]. SLP weights intensify when reaching the Indonesia/Borneo region from 150° to 180° phase. While the latter could produce a feedback onto the eastern Indian Ocean, the opposite phase of SLP variability originates in the Southern Ocean.

3. Discussion and Conclusion

[10] An MTM-SVD analysis of SST and SLP variability in the Indian Ocean over the 20th Century is conducted. It reveals significant climate signals therein and their spatio-temporal evolution are compared. No evidence existed for the QDO signal near 11-year period, as observed elsewhere [White and Tourre, 2003]. This could be due to the relatively small size of the Indian Ocean and deserves further investigation. However, the BDO signal, the QBO signal of 2.2-year period and three ENSO signals of 3.4-, 4.4, and 6.2-year period are observed in the Indian Ocean, as in the Global Ocean [White and Tourre, 2003]. The QBO and ENSO signals share similar evolution with similar physical mechanisms to maintain their amplitudes against dissipation.

[11] The spatio-temporal evolution of BDO, ENSO, and QBO signals in the Indian Ocean are preceded by SLP anomalous patterns in the Southern Ocean propagating northward and eastward from Madagascar to the eastern tropical Indian Ocean. This corroborates results by White and Tourre [2003]. This is indicative of the influence that climate variability in the Southern Ocean exerts on tropical variability in the Indian Ocean [White et al., 2002]. Another source of influence are the global equatorial coupled waves entering the western equatorial Indian Ocean [e.g., White and Tourre, 2003]. When this occurs, SST variability blooms there, from the Horn of Africa and continues to propagate eastward, over the next 180° of phase. In these phase sequences, there is little evidence of the IOD. A weak dipole-like does exist in the SST pattern near 90° or 120° phases (Figure 2). But this structure is short lived, overwhelmed by the equatorial coupled waves propagating eastward across the equatorial Indian Ocean and continuing across the equatorial Pacific Ocean [White and Cayan, 2000; White et al., 2002; White and Tourre, 2003; White and Annis, 2003].

[12] Global coupled waves associated with QBO and ENSO signals are referred to as global biennial wave (GBW) and global ENSO wave (GEW) [e.g., White and Cayan, 2000]. Their coupling includes anomalous SST-induced atmospheric circulation anomalies driving anomalous SST tendency (e.g., via anomalous wind-driven heat advection, mixing, and/or sensible-latent heat flux) overhead and displaced to the east. SST anomalies maintained against dissipation propagate eastward. Determining the specific oceanic heating processes that account for the QBO and ENSO signals will lead to better understanding and prediction of climate across the Indo-Pacific basin.