GEOPHYSICAL RESEARCH LETTERS, VOL. 36, L08702, doi:10.1029/2009gl037982, 2009 How rare are the 2006 2008 positive Indian Ocean Dipole events? An IPCC AR4 climate model perspective W. Cai, 1 A. Sullivan, 1 and T. Cowan 1 Received 4 March 2009; revised 4 March 2009; accepted 19 March 2009; published 18 April 2009. [1] The occurrence of three consecutive positive Indian Ocean Dipole (piod) events through 2006 2008, along with the unusual piod-la Niña combination in 2007, calls for a statistical assessment into the rarity of such events. To this end, we take 50 years from 19 climate models submitted for the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment (AR4) to provide a 950-year realization. Only six models produce a total of nine piod-la Niña pairs, with one occurrence of three consecutive piod events. The rarity is not due to an overly strong model El Niño-pIOD coherence. Two models produce the 2006 2007 sequence of one piod with an El Niño followed by a piod in a La Niña year, however, no models simulate the observed mechanism. Although the triggering processes vary between models, a commonality is that the trigger resides in the ocean, highlighting the importance of subsurface ocean observations in predicting piods. Citation: Cai, W., A. Sullivan, and T. Cowan (2009), How rare are the 2006 2008 positive Indian Ocean Dipole events? An IPCC AR4 climate model perspective, Geophys. Res. Lett., 36, L08702, doi:10.1029/2009gl037982. 1. Introduction [2] A piod event refers to a phase with low sea surface temperature (SST) anomalies in the eastern Indian Ocean (IO) and warm anomalies in the west [Saji et al., 1999; Webster et al., 1999]. During 2006 2008, the IO experienced a rare occurrence of three consecutive positive dipole events. Furthermore, the 2007 event occurred in a La Niña year, which had El Niño-Southern Oscillation (ENSO)- induced anomalies unfavourable for the development of such events. Likewise, during the 2008 piod development phase, the Pacific condition is La Niña-like, although not clear if it can be classified as a La Niña. In general, piod events contribute to droughts in East Asia, Australia, the Arabian Peninsula and flooding to parts of India and East Africa [Saji et al., 1999; Yamagata et al., 2004]. Indeed, the 2006 2008 consecutive piods have exacerbated the recent drought in south eastern Australia, making it the most severe (during 2001 2008) in over 100 years. According to the Meyers et al. [2007] classification of piods, the only precedence in the instrumental record for a consecutive triple piod series occurred during 1944 1946, when the Pacific experienced neutral conditions, contributing to the 1 CSIRO Marine and Atmospheric Research, Aspendale, Victoria, Australia. Copyright 2009 by the American Geophysical Union. 0094-8276/09/2009GL037982 most severe drought in Australia recorded to that date. Based on the Meyers et al. [2007] classification, the piod-la Niña pair in 2007 is unique, although according to Behera et al. [2008], a piod and La Niña pair occurred in 1967. [3] The surface circulation anomalies associated with the 2007 piod and La Niña pair are documented using available surface observations and outputs of a model that predicts such events [Luo et al., 2008]. The combination means a three-cell structure of the Walker Circulation is present in the tropical Indo-Pacific system, different from the usual two-cell structure, when a piod occurs in an El Niño year. The associated subsurface ocean circulation anomalies of the three consecutive piods are depicted in a study by Cai et al. [2009] using the newly available Argo data. They reveal that a shallowing thermocline signal curves sharply to the equator without going through a western boundary (WB) reflection. Once the thermocline anomaly reaches the equator and induces equatorial Kelvin waves, easterly anomalies ensue. This process initiates the 2007 piod event. The role of the easterly anomalies thereafter has been discussed by Luo et al. [2008]. [4] How rare is the occurrence of a piod-la Niña pair in the same year? How frequent is the occurrence of three consecutive piod events, as observed in 2006 2008? We examine 20th century experiments from 24 available IPCC AR4 climate models to provide the statistics, and describe, when present, the associated mechanism. 2. Model Data and IOD Definition [5] We take a 50-year (1950 1999) sample of one experiment from each of the 24 models to provide a large inter-model space of multi-century realizations. The detrended outputs of temperature, wind and thermocline (depth of the 20 C isotherm) are interpolated onto a common grid (0.8 1.9 ). We focus on the September, October, November (SON) season, when an IOD matures. To benchmark model performance, we use an updated version of the observed Global Sea Ice and Sea Surface Temperature [Rayner et al., 2003]. [6] In our study, the IOD is represented by the leading component of an Empirical Orthorgonal Function (EOF) analysis on SON SST anomalies. This definition does not impose the dipolar structure and allows each model to show its own IOD-like pattern [Saji et al., 2006]. Such EOF analysis is carried out for all 24 models and the observed in the tropical IO domain (40 E 120 E, 25 S 25 N). The variance is expressed in the EOF pattern, and the associated time series, taken as our IOD index, is scaled so that the standard deviation is one. We refer to an IOD event when L08702 1of5
statistically significant at the 95% confidence level. More models generate a coherence that is lower than the observed. The overall low coherence should be favourable for the generation of piod-la Niña occurrences. Thus, any rarity of the piod-la Niña pair is not attributable to an overly strong control of ENSO over the IOD. Figure 1. (a) Observed SON EOF1 pattern of tropical IO SST anomalies taken as the IOD, compared with (b) the all-model average. the amplitude exceeds one-standard deviation value. Likewise, we refer to an ENSO event when the normalised SON Niño3.4 index exceeds one standard deviation. Although the all-model average piod pattern shows a stronger eastwest gradient than the observed (Figure 1), the structure is reasonably realistic. However, significant inter-model variations exist. [7] An examination identifies five models in which the IOD pattern is rather unrealistic. These models are ECHO-G, IPSL-CM4, GISS-AOM, GISS-ER, and INM-CM3.0. Their patterns are characterized by weak but scattered anomalies, dissimilar to the observed. We conduct an objective test based on the global anomaly pattern associated with SON eastern IO SST anomalies (averaged over 90 E 110 E, 5 S 15 S) through a simple regression. An all-model average global pattern is constructed to represent the best model pattern, due to the cancellation of biases. Figure 2a plots the standard deviation of east IO SST anomaly time series versus the pattern correlation between the all-model average global pattern and the global pattern for each model. In the five models, the associated pattern is considerably different from the all-model average, as indicated by a low or negative correlation. We exclude these five models, reducing our model sample size to 19 with a 950-year realization. 3. ENSO-IOD Relationship [8] In order to examine the frequency of piod-la Niña occurrences, an assessment of the ENSO-IOD relationship is in order (Figure 2b). Using 17 models, Saji et al. [2006] showed that there is no systematic relationship between the IOD and ENSO. Figure 2b plots inter-model variations of the ENSO amplitude (standard deviation of Niño3.4) against inter-model variations of an ENSO-IOD correlation, together with that from the observed. The relationship, based on the 19 models, with a correlation of 0.39, is not 4. Occurrences of a piod-la Niña Pair [9] The 19 models EOF1 time series are examined to identify piod-la Niña occurrences. Only six models produce such events and these are shown in Figure 2b (star pattern). All six models produce an IOD-ENSO correlation lower than the observed. Their time series highlight several features (Figure 3). Firstly, out of the 950-year sample, there are only 9 such events, i.e., less than 1% of the years. Secondly, there are only five occurrences of two consecutive piods, with each pair s second piod in a La Niña year. Out of the five pairs, only two (CNRM-CM3 and CSIRO-Mk3.0) follow the observed sequence, i.e., one piod in an El Niño year followed by a piod in a La Niña year. Only CNRM- CM3 produces a mechanism similar, but not identical, to that of the 2007 piod, as will be discussed in section 5. Finally, there is only one occurrence of consecutive triple piods (in MIROC3.2(medres)). [10] If we relax the definition and count events that exceed 0.75 of the standard deviation in both the IOD index and Niño3.4, the total number of piod-la Niña occurrences increase to 19, or about 2% of the years. The majority (16) still occur in the six models discussed above (Figure 3). Two of the additional events are produced in BCCR-BCM2, and Figure 2. (a) Standard deviation of east IO SST anomaly time series (averaged over 90 E 110 E, 5 S 15 S, in C) versus pattern correlation between a global pattern associated with the time series and the all-model averaged pattern. The global pattern for each model is obtained by regressing grid-point global SST anomalies onto the time series. (b) ENSO amplitude ( C) versus correlations between Niño3.4 and the IO SST EOF1 time series (i.e., the IOD). Models with a star pattern are ones with piod-la Niña occurrences. 2of5
Figure 3. Time series of SON SST EOF1 over the tropical IO region (blue bars) and SON Niño3.4 (line with dots) in six models that produce piod-la Niña occurrences. Single piod event, or consecutive piod events with one coinciding with alaniña event are shown by green bars, and the concurrent ENSO index is indicated by highlighted circles, respectively. All data are normalised, meaning that exceeding one standard-deviation is classified as an IOD or ENSO event. one in PCM1. The occurrences of three consecutive piods increases to two, the addition produced by PCM1, all in non- La Niña years. 5. Role of Ocean Dynamics in Driving the piod- La Niña Combination [11] Usually, an initial anomaly of negative SST, suppressed rainfall, or easterlies off the Java coast provokes a reduction in heating of the atmospheric column. The baroclinic atmospheric response includes an increased sea level pressure and an anticyclonic circulation that is consistent with the response (Gill model) to a heat sink displaced south of the equator. Because the Sumatra-Java coast is tilted from the southeast to the northwest, upwelling is promoted on the eastern side of the anticyclone, while the near equatorial easterly anomalies elevate the thermocline to the east. The consequence is further cooling and a suppression of rain in the Java upwelling zone. This positive feedback involving anomalous SST, rainfall, winds, and thermocline depth in the east occurs in May October when the mean thermocline in the east is shallow and seasonal upwelling occurs along the coast [Saji et al., 1999; Webster et al., 1999]. Besides El Niño-induced anomalies other possible triggers may include the Southern Annular Mode [Lau and Nath, 2004] and the onset of the Asia monsoon [Fischer et al., 2005]. As it is a positive feedback process, it may be initiated by a perturbation in winds, SST or the thermocline in the Java-upwelling zone. [12] One can rule out El Niño-induced easterly anomalies as a trigger, because of the conjunction with La Niña. Apart from this atmospheric trigger, off-equatorial upwelling Rossby waves, upon their WB reflection into equatorial upwelling Kelvin waves, can also be a piod trigger. Equatorial IO easterly anomalies, while generating piod-inducive upwelling Kelvin waves, also generate an off-equatorial Ekman ridge. The ridge propagates westward as downwelling Rossby waves and reflects as equatorial downwelling Kelvin waves, which may trigger a negative IOD. This is a process for the biennial cycle of the IOD [Feng and Meyers, 2003]. In an opposite manner, off-equatorial upwelling Rossby waves, however induced, may lead to a piod. [13] For the 2007 piod-la Niña pair, Argo reveals that an off-equatorial downwelling Rossby wave is generated by equatorial easterly anomalies associated with the 2006 El Niño [Cai et al., 2009]. Its reflection as an equatorial downwelling Kelvin wave should induce a negative IOD event in 2007. Instead, a negative thermocline anomaly off the Java upwelling zone associated with the 2006 piod radiates into the interior IO. Supported by the wind stress curl, it curves sharply equatorward, to arrive at the equator, initiating piod-conducive easterly anomalies along the equator, leading to the 2007 piod event. The process that generates the wind stress curl reinforcing the "equatorwardcurving" process of thermocline anomalies is not clear. [14] We examine all the piod events shown in Figure 3 by inspecting the monthly evolution of the thermocline, SST and wind anomalies, in terms of Rossby wave reflection at the WB, the equator curving process described above, or other perturbations. The results are summarised in Table 1. Many are triggered by the WB mechanism. However, in the CSIRO-Mk3.0, most IODs are driven by an unrealistic oceanic teleconnection in that the entire Sumatra-Java region participates in the El Niño-discharge (E-D): Pacific Rossby waves associated with the El Niño discharge propagate to the model Java coast, triggering an IOD event about 6 9 months after the matured ENSO [Cai et al., 2005]. [15] It happens that in several models listed in Table 1, the eastern half of the Sumatra-Java water participates in the E-D process. This is rather unrealistic, and appears to be 3of5
Table 1. Climate Models That Produce Single and Consecutive piod Events With La Niña From 1950 1999 a Models IOD-ENSO Cor.: Zero (1-Year) Lag 1st piod: ENSO 2nd piod: ENSO 3rd piod: ENSO Single piod With La Niña CCSM3 0.13 ( 0.01) N (WB) L (WB) CGCM3.1(T63) 0.00 ( 0.02) L (WB) L (WB) CNRM-CM3 0.06 (0.75) E (E-D) L (E-D) 2 (E-D), (E-D) CSIRO-Mk3.0 0.14 (0.70) E (E-D) L (E-D) MIROC3-2(medres) 0.30 (0.57) E (WB) N (E-D) L (WB) UKMO-HadCM3 0.56 (0.28) L (E-D) a The mechanisms for the piod development include western boundary (WB) reflection of off-equatorial Rossby waves, equatorward-curving process of a negative thermocline anomaly to the equator, and negative anomalies associated with an El Niño-discharge (E-D). La Niña, El Niño, and Neutral conditions are indicated by L, E, and N respectively. associated with the model ENSO anomaly pattern that extends too far west. Such a bias is indicated by the strong IOD-ENSO correlation at the one-year lag (ENSO leads, see Table 1). For example, in CNRM-CM3, although there is virtually no correlation at zero lag, the correlation at the 1-year lag is 0.75. Similarly, in MIROC3-2(medres), the 1-year lag correlation exceeds the zero lag. Further study is needed to unravel the cause of the unrealistic lag in these models. [16] Here we focus on UKMO HadCM3, in which E-D plays a part in the piod generation (Figures 4a 4d). The process of E-D off the eastern half of the Sumatra-Java coast persists into mid-austral winter after an El Niño peaks, elevating the thermocline and generating a SST gradient and easterly anomalies, which consolidate the shoaling thermocline. In doing so, it prevents the eastward-extending equatorial downwelling Kelvin wave from creating a negative IOD. Another example, is CNRM-CM3 (Figures 4e 4h), in which a similar E-D process operates. Initially, the discharge signal, which is seen to propagate from the western Pacific, is situated away to the southwest of the Sumatra-Java coast; however, once wind anomalies are generated, the centre of cold anomalies move immediately off the coast, as a consequence of the anomalous coastal upwelling. [17] Thus, although the process varies from one model to another with an unrealistic E-D process, a commonality is that the trigger resides in the ocean, reinforcing the role of the ocean in generating piod-la Niña events. This is not Figure 4. Evolution of anomalous 20 C isotherm depth (m) and wind stress (vectors, N m 2 ) in the IO domain, (a d) associated with a single piod-la Niña event from May-August, based on UKMO HadCM3. (e h) A single event in the CNRM-CM3 model, covering May-November at a two-month interval. Max vector size = 0.06 N m 2. 4of5
surprising given that the La Niña-induced atmospheric anomalies are unfavourable for piod development. 6. Conclusions [18] Outputs of wind, SST and thermocline depth from climate models submitted for the IPCC AR4 are used to address the rarity of the observed 2006 2008 piod events. An examination identifies 19 models which produce reasonable IOD-like patterns, providing a 950-year realization. We define IOD and ENSO events as surpassing one-standard deviation of their respective indices. Only six models produce a total nine piod-la Niña occurrences, or less than 1% of the years. There is only one occurrence of three consecutive piods. The model rarity is not due to an overly strong ENSO control over IO variability, because most models produce an ENSO-IOD coherence lower than the observed. Only two models produce the observed sequence of one piod with an El Niño followed by one with a La Niña, similar to the observed events of 2006 and 2007. No model produces the exact mechanism as in the observations, which may be in part due to the limitations of the models temporal resolutions. A commonality of such events is that the triggering mechanisms reside in the ocean. This is not surprising given that the atmospheric anomalies from the Pacific are unfavourable for piod development, however it does highlight the importance of subsurface ocean observations in predicting such occurrences. [19] Acknowledgments. W. Cai, A. Sullivan, and T. Cowan are supported by the Australian Department of Climate Change and the Wealth from Oceans National Research Flagship. We are grateful to the Associate Editor and two reviewers for their helpful comments, which improved the manuscript. We also acknowledge the significant work that the Program for Climate Model Diagnosis and Intercomparison (PCMDI) has undertaken in collecting and storing the model data http://www-pcmdi.llnl.gov). References Behera, S. K., J.-J. Luo, and T. Yamagata (2008), Unusual IOD event of 2007, Geophys. Res. Lett., 35, L14S11, doi:10.1029/2008gl034122. Cai, W., H. H. Hendon, and G. Meyers (2005), Indian Ocean Dipole-like variability in the CSIRO Mark 3 coupled climate model, J. Clim., 18, 1449 1468. Cai, W., A. Pan, D. Roemmich, T. Cowan, and X. Guo (2009), Argo profiles a rare occurrence of three consecutive positive Indian Ocean Dipole events 2006 2008, Geophys. Res. Lett., doi:10.1029/ 2008GL037038, in press. Feng, M., and G. Meyers (2003), Interannual variability in the tropical Indian Ocean: A two-year time scale of Indian Ocean Dipole, Deep Sea Res., Part II, 50, 2263 2284. Fischer, A., P. Terray, E. Guilyardi, S. Gualdi, and P. Delecluse (2005), Two independent triggers for the Indian Ocean Dipole zonal mode in a coupled GCM, J. Clim., 18, 3428 3449. Lau, N.-C., and M. J. Nath (2004), Coupled GCM simulation of atmosphereocean variability associated with the zonally asymmetric SST changes in the tropical Indian Ocean, J. Clim., 17, 245 265. Luo, J.-J., S. Behera, Y. Masumoto, H. Sakuma, and T. Yamagata (2008), Successful prediction of the consecutive IOD in 2006 and 2007, Geophys. Res. Lett., 35, L14S02, doi:10.1029/2007gl032793. Meyers, G. A., P. C. McIntosh, L. Pigot, and M. J. Pook (2007), The years of El Niño, La Niña, and interactions with the tropical Indian Ocean, J. Clim., 20, 2872 2880. Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan (2003), Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century, J. Geophys. Res., 108(D14), 4407, doi:10.1029/ 2002JD002670. Saji,N.H.,B.N.Goswami,P.N.Vinayachandran,andT.Yamagata (1999), A dipole mode in the tropical Indian Ocean, Nature, 401, 360 363. Saji, N. H., S.-P. Xie, and T. Yamagata (2006), Tropical Indian Ocean variability in the IPCC twentieth-century climate simulations, J. Clim., 19, 4397 4417. Webster, P. J., A. M. Moore, J. P. Loschnigg, and R. R. Leben (1999), Coupled ocean-atmosphere dynamics in the Indian Ocean during 1997 98, Nature, 401, 356 360. Yamagata, T., S. K. Behera, J.-J. Luo, S. Masson, M. Jury, and S. A. Rao (2004), Coupled ocean-atmosphere variability in the tropical Indian Ocean, in Earth s Climate: The Ocean-Atmosphere Interaction, Geophys. Monogr. Ser., vol. 147, edited by C. Wang, S.-P. Xie, and J. A. Carton, pp. 189 212, AGU, Washington, D. C. W. Cai, T. Cowan, and A. Sullivan, CSIRO Marine and Atmospheric Research, PMB 1, Aspendale, Vic 3195, Australia. (wenju.cai@csiro.au) 5of5