Influence of the Indian Ocean Dipole on the Australian winter rainfall



[1] Using an atmospheric general circulation model and observed datasets of sea surface temperature and rainfall, we studied the influence of the Indian Ocean Dipole (IOD) on the Australian winter rainfall. The IOD has significant negative partial correlations with rainfall over the western and southern regions of Australia. These negative partial correlations extend south-eastward from Indonesia all the way to south east Australia. Our atmospheric general circulation model sensitivity experiments indicate that cold sea surface temperature anomalies prevailing west of the Indonesian archipelago during the positive IOD events introduce an anomalous anticyclonic circulation at lower levels over the eastern tropical and subtropical Indian Ocean, and over much of the Australian continent. It is also apparent that the response of the atmosphere to the IOD in this region is baroclinic, causing anomalous subsidence and anomalous reduction in the rainfall over the affected regions of Australia.

1. Introduction

[2] Winter crops and/or animal husbandry can be found in many parts of Australia, such as the southwest region of the state of Western Australia, the southern region of the state of South Australia, and southern Victoria. The percentage of annual rainfall that can be attributed to the rainfall during the winter months of June–September (JJAS), based on the analysis of the CMAP rainfall data [Xie and Arkin, 1996], is presented in Figure 1. There have been several studies indicating that sea surface temperatures influence the Australian winter rainfall. Streten [1981, 1983] claimed that years of extensive Australian droughts are linked to the negative sea surface temperature anomalies (SSTA) over the eastern Indian Ocean. Nicholls [1989] found strong negative correlations between the first rotated principal component of Australian winter rainfall and the sea surface temperatures in the central Indian Ocean, with the strongest negative correlation values centered in the box of 10°–20°S, 80°–90°E. He found positive correlations around Indonesia, in agreement with earlier studies [Streten, 1981, 1983; Whetton, 1986]. Nicholls [1989] also found that the first rotated principal component of Australian winter rainfall has a correlation coefficient of 0.75 with the SST difference between the Indonesian region (0°–10°S, 120°E–130°E) and the central Indian Ocean (10°–20°S, 80°–90°E), after the ENSO related component was removed from the SST through partial correlation technique. He showed that this SST difference influences the winter rainfall over a broad area of Australia, stretching from northwest to southeast.

Figure 1.

Percentage of the winter (JJAS) rainfall in the annual rainfall.

[3] The recently discovered Indian Ocean Dipole (IOD) mode [Saji et al., 1999; Webster et al., 1999] has considerable influence on the climates of many regions around the globe [Saji et al., 1999; Ashok et al., 2001; Guan and Yamagata, 2003; N. H. Saji and T. Yamagata, Interference of teleconnection patterns generated from tropical Indian and Pacific Oceans, submitted to Climate Research, hereinafter referred to as SY, 2003]. Whenever the Indian Ocean Dipole Mode Index (IODMI), defined as the area-averaged SST anomaly difference between the tropical western Indian Ocean (50°E–70°E, 10°S–10°N) and the tropical southeastern Indian Ocean (90°E–110°E, 10°S-equator), is positive, it leads to droughts over the Indonesian region, and heavy rains and floods over East Africa [Saji et al., 1999]. When the sign of the IODMI reverses, these anomalous responses also swing to the opposite phase. SY, 2003 demonstrate that the IOD has an impact on the winter climate of the southern hemisphere during June–October.

[4] It may be noted that Nicholls [1989] spatial pattern extends from the tropical Ocean surrounding the Indonesia to the central Indian Ocean. IOD, on the other hand, extends farther west, all the way into the western tropical Indian Ocean. The western (eastern) box of Nicholls' spatial pattern also extends farther south (east) in comparison to the IOD. By definition, Nicholls' dipole-like SSTA pattern extends into the tropical Pacific, while the areal extent of the IOD is confined to the tropical Indian Ocean.

[5] Motivated by the aforementioned research that brought out the impact of the recent strong IOD events on the climates of many regions of the globe, in this paper, we study the influence of the IOD on the Australian winter rainfall. We use an atmospheric general circulation model (AGCM) as well as observational datasets to understand the relationship and the mechanisms behind.

2. Data, Model, and Methodology

[6] The CMAP precipitation data [Xie and Arkin, 1996] from 1979–1997, derived from satellite measurements and ground observations, have been used in this study. We used the GISST 2.3b dataset [Rayner et al., 1996] for the same period to compute the IODMI index, and NINO3 Index (obtained by area-averaging the sea surface temperatures over 5°N–5°S, 150°W–90°W) that is widely used as an index of El Niño/Southern Oscillation (ENSO). NCEP/NCAR reanalysis data [Kalnay et al., 1996] has also been used. We used the Frontier Atmospheric General circulation model version 1.0 (FrAM1.0) AGCM [Guan et al., 2000; Ashok et al., 2001; Guan and Yamagata, 2003; K. Ashok et al., On the individual and combined influences of the ENSO and the Indian Ocean Dipole on the Indian summer monsoon, submitted to Journal of Climate, hereinafter referred to as KA, 2003] to carry out two sensitivity experiments to assess the IOD influence on Australia. The model is a spectral model with T42 resolution in the horizontal direction, with 28 levels in the vertical direction, and has full physics. The aforementioned publications may be referred to for the full details of the model. In these experiments, the model has been integrated for a full calendar year starting from 1 January. Each experiment is an ensemble of sixteen realizations that differ from one another in the initial conditions. To obtain these initial conditions, we integrated the model for 20 calendar years with seasonally varying SST as the lower boundary condition; we selected the 1 January conditions of the last 16 years of this long run as the initial conditions for the member ensembles in the present study. In the first experiment, referred to as the control experiment, we have imposed the seasonally varying climatological SST as the lower boundary condition. In the second experiment, we have imposed positive IOD type of SSTA (Figure 2), obtained by correlation analysis, on the lower boundary condition used in the control experiment. The SSTA in the last experiment is imposed from the month of April, and its amplitude increases as observed [Saji et al., 1999]. A similar procedure was adopted in AGCM sensitivity studies that assessed the influences of the IOD and ENSO on climates of different regions [Ashok et al., 2001; Guan et al., 2003; KA, 2003]. As an example, the SSTA imposed during the month of September in the second experiment is presented in Figure 2. These anomalies are canonical in nature i. e. they do not represent the SSTA for any specific positive IOD events; however, the magnitude, and the distribution in general, are comparable to those observed in the Indian Ocean during strong IOD events such as 1961, 1994, 1997 etc. The results from the control experiment are subtracted from those from the positive IOD experiment to obtain the various anomalies due to the imposed IOD SSTA in the latter experiment.

Figure 2.

The SSTA (°C) imposed in the positive IOD experiment for September.

3. Results

[7] We have carried out partial correlation analysis for the winter season between the IODMI and the rainfall anomalies, while considering the NINO3 index as the second predictor. The rationale for removing the ENSO influence is based on the fact that many of the IOD and ENSO events co-occur together, and hence the use of simple correlation technique artificially masks/amplifies each other's influence depending upon the region of influence [Panofsky and Brier, 1958; Ashok et al., 2003; SY, 2003]. A somewhat similar procedure has been followed in different studies that examined the role of the Indian Ocean SSTA on global climate [Nicholls, 1989; Guan et al., 2003; SY, 2003; KA, 2003]. Positive correlations can be found in the tropical western Indian Ocean (Figure 3a). Significant negative correlations (magnitude greater than 0.38 at 90% confidence level for the 20-year data) can be seen over the Indian Ocean off Indonesia, south eastern Indian Ocean, southern regions of the state of west Australia, parts of South Australia, and Victoria, in agreement with SY, 2003. Though the correlations do not appear to be very strong everywhere over Australia, nonetheless, intense IOD events can have a very prominent influence over Australian winter rainfall, just as they modulate Indian summer monsoon [Ashok et al., 2001]. As an example, we show the normalized rainfall anomalies over Australia during the winter of 1994 (Figure 3b), a year when strong IOD event occurred. Large rainfall deficits occurred over southwest and southern Australia. The whole region extending from the west coast of Indonesia towards Australia - and Australia as a whole - experienced negative rainfall anomalies during this year. The distributions of the rainfall anomalies of other ‘pure’ positive IOD events (when there is no co-occurring El Niño) such as those of 1961, and 1967, obtained using NCEP-NCAR reanalysis data, exhibit large deficits, similar to that of 1994, over the Australian continent (figures not shown). The rainfall anomalies are similar in many other positive IOD years such as 1972, 1982, 1983, 1997 etc. (figures not shown).

Figure 3.

(a) Partial correlations between the IODMI and the rainfall during JJAS (1979–1997). Negative correlations significant at 90% confidence level (>0.38) are shaded. (b) Normalized rainfall anomalies during JJAS, 1994. Positive values are shaded.

[8] The simulated rainfall anomalies induced by the positive IOD type of SSTA have the same sign in general as that of the observed rainfall anomalies during the positive IOD years (Figure 4a; compare with Figure 3a). There is a rainfall deficit center in the tropical eastern Indian Ocean anomaly off the coast of Indonesia. The negative rainfall anomalies envelop a large area that includes almost all of Australia. The negative rainfall anomalies observed over the tropical western Pacific are also well simulated. The rainfall anomaly distribution over the western pole of the Indian Ocean Dipole is realistic. At 850 hPa (200 hPa), the anomalous wind circulation over tropical south eastern Indian Ocean is anticyclonic (cyclonic), implying that the atmospheric response to the IOD in the tropics and subtropics is baroclinic (Figures 4a and 4b), in agreement with earlier studies [Ashok et al., 2001; Guan et al., 2003].

Figure 4.

(a) Simulated JJAS rainfall anomalies (mm·d−1) and anomalous winds (m·s−1) at 850 hPa. Significant rainfall anomalies (at 90% confidence level from the 2-tailed student's t-test) are shaded. (b) Anomalous winds (m·s−1) at 200 hPa.

[9] To understand the mechanism behind the impact of the IOD on the Australian winter climate, we have calculated the difference of the velocity potential field by subtracting the JJAS mean velocity field simulated in the control experiment from that of the experiment in which the IOD SSTA were imposed (Figures 5a and 5b). It is seen that at 850 hPa there is a zone of convergence in the western Indian Ocean, flanked by the anomalous divergence to the west (Figure 5a). The western tropical and subtropical Indian Ocean, the western tropical Pacific and Australian regions are under the anomalous divergence zone. The ridge at 5°S, south west off Indonesia, is due the cold temperature anomalies imposed there. It should also be noted that, when an ENSO event co-occurs with a positive IOD event, the divergence in western Pacific is much more amplified (KA, 2003). The distribution of the velocity potential anomalies at 200 hPa confirms that the atmospheric response is baroclinic (Figure 5b). At this level, part of the divergent anomalies from the western pole converges directly over to Australia. At lower levels, on the other hand, some of the anomalous divergent winds emanating from the Australian continent are shown to converge over the warmer western pole (Figure 5b). This means that the lower level convergence over the western pole, apart from enhancing the low level divergence over the eastern pole, also directly induces anomalous divergence over Australia, and also west Pacific. Similarly, at 200 hPa, the anomalous divergence from the western pole, apart from converging over the eastern pole, also induces a zone of anomalous convergence over Australia and Western Pacific. Thus, during a positive IOD event, both the warmer western pole, and colder eastern pole by virtue of its immediate presence near the western Pacific and Australia, modulate the zonal circulation over a wide region.

Figure 5.

Simulated anomalous velocity potential (m2·s−1) and anomalous divergent winds (m·s−1) (a) at 850 hPa (b) 200 hPa. (c) Sea level pressure anomalies (hPa) in contours, and vertically integrated moisture flux anomalies (in vectors; 10−3 kg·m−1·s−1). Positive values are shaded.

[10] We have also presented the anomalous sea level pressure (SLP) distribution along with that of the vertically integrated moisture flux anomalies in Figure 5c. Over the warmer western pole, we see an anomalously low SLP center, flanked by a high pressure anomaly center to the east. The SLP anomaly pattern shown over the IOD and Australian regions is, in general, typical of the positive IOD years. The simulated sea level pressure anomalies conform to the simulated circulation and rainfall features. The vertically integrated moisture flux anomaly pattern shows basically a transfer of moisture from the eastern Indian Ocean towards the western Indian Ocean.

4. Summary and Conclusion

[11] In this paper, the impact of the IOD on the Australian winter rainfall was studied using different observational data sets and the FrAM 1.0 AGCM. The results from this study can be summarized as follows:

[12] During the intense IOD years such as 1961, 1967, 1972, 1994 and 1997 etc., large negative rainfall anomalies are observed over many parts of the southern half of the Australian continent. We have computed partial correlations between the IODMI and the rainfall over Australia after removing the influence of the NINO3 index, the second predictor that represents the ENSO events in the Pacific. We found that the IOD has significant impact on the winter rainfall of western and southern Australia. Significant negative partial correlations extend southeastward from Indonesia all the way to southeast Australia. Using the FrAM 1.0 AGCM, we conducted two sets of sensitivity experiments to understand the mechanism behind the IOD-Australian winter rainfall relationship. The simulations reasonably agree with observations. The mechanism behind how the positive IOD events influence the Australian winter climate can be derived from the results of these experiments: The relatively cold SSTA to the west of the Indonesian archipelago during a positive IOD event introduce an anomalously anticyclonic circulation at lower levels over the eastern tropical and subtropical Indian Ocean as well as much of the Australian continent, consistent with the Matsuno-Gill theory [Matsuno, 1966; Gill, 1980; Wang et al., 2003; Guan et al., 2003]. The response of the atmosphere in this region is baroclinic. This results in anomalous subsidence over regions of western and southern Australian continent and also reduces the rainfall.

[13] The present study, along with earlier studies that reveal the IOD impact on global climate [Behera et al., 1999; Saji et al., 1999; Ashok et al., 2001; Guan et al., 2003; Guan and Yamagata, 2003; Zubair et al., 2003; SY, 2003; KA, 2003] provides an impetus to use the IODMI as a diagnostic tool in global teleconnection studies. In this background, it is also imperative to develop forecasting techniques to predict the IOD indices that, in turn, can be used in the prediction of seasonal weather over the relevant regions of the globe.


[14] The authors thank Prof. D. V. Bhaskar Rao, Drs. S. K. Behera, W. L. Chan, and S. A. Rao for discussions on the present work. The two anonymous reviewers are thanked for their comments, which have improved the paper. Figures presented in this work have been prepared using the GrADS software.