La Niña Modoki impacts Australia autumn rainfall variability



[1] Both El Niño-Southern Oscillation (ENSO) and ENSO Modoki affect Australian rainfall but the commonalities and contrasts of their impacts have not been fully explored. We show that both types feature a strong asymmetry between impacts of La Niña and El Niño in austral autumn (March–May); the La Niña-Australian rainfall teleconnection is statistically significant, whereas the El Niño-Australian rainfall relationship is not. A La Niña Modoki cold anomaly near the Dateline is effective in shifting convection westward, causing an autumn rainfall increase over northwestern Australia extending to the northern Murray-Darling Basin, rather than over the east as in a conventional La Niña. During an El Niño Modoki, the tendency for lower Australian rainfall is far weaker. The asymmetry explains the strong inter-ENSO variations in rainfall anomalies, including 1983, when a strong El Niño residual was associated with a wet autumn. Our results highlight the importance of considering the influence from La Niña Modoki in predicting ENSO's impacts.

1. Introduction

[2] It is well known that ENSO exerts significant impacts on Australian rainfall [e.g., McBride and Nicholls, 1983]. During an El Niño-phase, Australia is thought to experience below-average rainfall, however this is only true in a statistical sense, as there are strong inter-El Niño variations in the teleconnection. For example, the 1997/1998 El Niño event was the strongest in terms of conventional oceanic indices such as NINO3 (sea surface temperature (SST) anomalies averaged over 5°S–5°N, 150°W–90°W), yet Australian rainfall was about average. In contrast, the weak 2002/2003 El Niño event was associated with a severe drought across much of Australia.

[3] One explanation for this is an inter-event difference in the El Niño SST anomaly pattern. The 1997/98 event has a maximum SST anomaly situated in the eastern equatorial Pacific, whereas in 2002/2003 the anomaly centre is located at the International Dateline. The latter represents a new type of El Niño event called El Niño-Modoki [Ashok et al., 2007]. The entity including its opposite phase, La Niña Modoki, is referred to as ENSO Modoki. Similar but not identical terms describing departures from conventional ENSO include “Dateline ENSO” [Larkin and Harrison, 2005], or “Trans Niño” oscillation [Trenberth et al., 2002]. During an El Niño Modoki event, there are two anomalous Walker Circulation cells in the troposphere, instead of the single-celled pattern of the conventional El Niño. The core rising branch of the double-celled Walker Circulation is located over the central equatorial Pacific, and the associated western descending branch is situated over Indonesia and northern Australia, therefore being more effective in suppressing Australian rainfall.

[4] An excellent example is provided by Wang and Hendon [2007] contrasting the impact of an El Niño Modoki event in 2002 with a traditional El Niño event in 1997, in the September, October, November (SON) season. Taschetto and England [2009] compare the impact of El Niño Modokis in autumn (March, April, May; or MAM) with the impact of the conventional El Niño in SON. An explanation is still lacking for inter-event differences in autumn rainfall. For example, in terms of NINO3, the highest value was recorded in 1983, when the Indo-Pacific was still experiencing significant anomalous warming from the 1982 El Niño event, yet Australia recorded its second wettest autumn since official records started in 1900. Did ENSO Modoki play a part? We explore the similarities and differences between ENSO and ENSO Modoki in terms of their impacts on Australian rainfall, and show that the differences are most conspicuous in autumn. We show that a strong non-linearity (asymmetry) exists in this season in the impact of ENSO and ENSO Modoki, similar to that using all month data for ENSO [Power et al., 2006]. As such, the impacts from ENSO and ENSO Modoki in autumn mainly manifest through a La Niña phase, but the spatial patterns of their impacts are vastly different.

2. Reanalysis and Observations

[5] An updated version of the Global Sea Ice and SST reanalysis [Rayner et al., 2003] is used to construct an ENSO Modoki index (EMI) and NINO3 index. Reanalyses from the National Centre for Environmental Prediction (NCEP) [Kalnay et al., 1996] and observed Australian rainfall since 1950, subjected to extensive quality control from the Australian Bureau of Meteorology Research Centre (BMRC), are used to explore the processes associated with the teleconnections to rainfall. We note that prior to the 1970s the data quality is not as high as that after this time. Monthly and seasonal anomalies of various fields, referenced to the climatological mean over the period since 1950, are used. We also use the Southern Oscillation index (SOI) supplied by the BMRC as a comparison to the SST-based indices. In our study, the EMI is defined following Ashok et al. [2007], i.e., EMI = [SSTA]A − 0.5 × ([SSTA]B + [SSTA]C), where the brackets represent the area-averaged SST anomalies for the regions A (165°E–140°W, 10°S–10°N), B (110°W–70°W, 15°S–5°N), and C (125°E–145°E, 10°S–20°N), respectively. An examination of the EMI seasonality shows that its standard deviation is highest in January (0.44°C) and lowest in May (0.36°C). As we are interested in variability, all anomalies are linearly detrended over the 58-year period.

3. Seasonality of the Impacts From ENSO Modoki

[6] We commence by comparing the correlation of the three indices with all-Australian rainfall using monthly anomalies (Figure 1a), starting from 1950, in a 13-year sliding window, and record the coefficient at the seventh year (i.e., first year is 1956). By definition, when the SOI is positive, it signifies a La Niña, whereas when NINO3 is positive, it indicates an El Niño. Thus, correlations with the SOI and with NINO3 share similar fluctuations but of opposite sign. Previous studies have shown that these fluctuations are a consequence of modulation by the Interdecadal Pacific Oscillation (IPO) [Power et al., 1999]. When the IPO is in a positive phase, its correlation with Australian rainfall is low, and vice-versa (Figure 1a). Several other interesting features emerge. The correlations with the EMI fluctuate coherently with either of the other two curves, suggesting that the IPO also modulates the ENSO Modoki-rainfall teleconnection in a similar manner. The correlation with the EMI over the full period (annual values over 1950–2007) is larger than that with NINO3 (−0.56 and −0.39 respectively), but both are smaller than that with the SOI (0.61), due to the fact that one pole of the SOI contains variability of mean sea level pressure over Australia, which is a surrogate of rainfall variations. These results seemingly suggest that the impact from ENSO-Modoki varies only slightly from that of conventional ENSO.

Figure 1.

Time series of correlations between ENSO and ENSO-Modoki climate indices and all-Australia rainfall using detrended annual (all months) and seasonal data since 1950 through a 13-year sliding window. Climate indices used are the EMI (black), SOI (red) and NINO3 (orange). A time series of the IPO is also shown in Figure 1a (blue), based on the definition described by Parker et al. [2007, Figure 3].

[7] However, Australian rainfall variability shows both a strong seasonality and regionality, dominated in summer (December, January, February; or DJF) by northern Australia rain, and in winter (June, July, August; or JJA) and spring (SON) by southern Australia rainfall. We therefore stratify the correlations into four seasons (Figures 1b1e). In DJF and SON (Figures 1b and 1e), the correlations with the EMI are again rather similar to the correlations with NINO3 most of the time. But this is not true for MAM or JJA.

[8] An analysis for each season through the regression of seasonal SST anomalies onto the corresponding seasonal EMI generates anomaly patterns showing that in DJF and SON, warm anomalies occupy the entire equatorial Pacific, although the maximum is near the International Dateline. By contrast, in MAM and JJA, the patterns are rather different from those associated with NINO3, and display a well-defined tri-polar structure in the SST anomaly pattern. In MAM, the correlation between all-Australia rainfall and the EMI for 1950–2007 is statistically significant (−0.53). This is not the case in JJA (−0.25, although significant since the 1970s [Ashok et al., 2007]). We therefore focus our attention on the MAM season.

4. Impacts on Autumn Rainfall, ENSO Modoki Versus Conventional ENSO

[9] The above analysis suggests that the difference between the impact from ENSO and ENSO Modoki on Australia-wide rainfall is sharpest in MAM. The correlation between the EMI and NINO3 in MAM is −0.03, meaning the two indices may be regarded as independent for the purpose of rainfall correlations. The MAM rainfall patterns associated with the EMI and with NINO3 are indeed very different (Figures 2a and 2b). The correlations of MAM rainfall with NINO3 are small and not statistically significant in most regions, whereas the correlations with the EMI are greater and statistically significant in many regions particularly northern Australia. The weak correlation with NINO3, meaning that there is little influence from ENSO in MAM, is somewhat surprising and seems contradictory to previous results.

Figure 2.

Maps of correlation of grid-point Australian rainfall (a) with the EMI and (b) with NINO3 using detrended MAM data. Areas within white curves are significant at the 95% confidence level. (c) Scatter plot of the MAM EMI against MAM all-Australia rainfall. The linear fit for data with a negative EMI value is superimposed, showing correlation and regression coefficients. A correlation coefficient of 0.38 is required for statistical significance at the 95% confidence level. (d) Same as Figure 2c, but for NINO3. The circled NINO3 outlier (large El Niño in 1983) is not included in the linear fit.

[10] Further investigations reveal that there is an asymmetry between impacts from La Niña Modoki and from El Niño Modoki, and between those from the conventional La Niña and El Niño, as illustrated in Figures 2c and 2d. For negative values of the indices, the correlation with the EMI and with NINO3 is 0.46 and 0.43 respectively, both exceeding 0.38 required for a significance at the 95% confidence level. The sensitivity of rainfall to the EMI is an increase of 78.5 ± 30.4 mm season−1 °C−1 and the sensitivity to NINO3 is comparable at 81.8 ± 31.7 mm season−1 °C−1. By contrast, for positive index values, the impacts are not statistically significant. In particular, for positive NINO3, if the outlier (circled, Figure 2d) is included the slope becomes slightly positive (not shown), which would mean that an El Niño tends to increase Australian rainfall in MAM. However, this is solely due to the outlier autumn of 1983; without it, there is little sensitivity. The outlier per se highlights the strong influence of a La Niña Modoki modulating the impacts of El Niño, as will be discussed later in section 6.

[11] Thus, conventional ENSO does affect Australian MAM rainfall, but the influence is only significant during La Niña. This asymmetry of ENSO's impact was discussed in previous studies [Power et al., 2006]. We show that the impact of ENSO Modoki similarly manifests during La Niña Modoki. Below we focus on La Niña and La Niña Modoki, to address whether their influence shares a similar spatial pattern in terms of a rainfall teleconnection over Australia in MAM.

5. Circulation Anomalies: La Niña Modoki Versus La Niña

[12] To this end, an identical analysis is conducted to obtain correlations with the two indices, as in Figures 2c and 2d, but using grid-point circulation anomalies with a negative value of the indices (e.g. La Niña and La Niña Modoki). Patterns of rainfall correlation are plotted in Figures 3a and 3d, where a negative correlation mean higher rainfall (blue colour), because it is associated with a higher negative EMI or NINO3. The impact of La Niña Modoki extends from northwest Australia to the Murray-Darling Basin region, whereas the influence of La Niña concentrates on northeastern Australia and much of South Australia. Overall, there is a sense that the impact of La Niña Modoki comes from northwest.

Figure 3.

Map of correlation coefficients of negative EMI values with anomalies of (a) rainfall, (b) SST, and (c) OLR. In Figure 3a, negative correlations (blue) refer to a rainfall increase as higher rainfall is associated with a higher negative EMI, whereas in Figures 3b and 3c, a positive correlation (blue) means an anomaly of cold SST and negative OLR (increased convection), respectively. (d–f) Same as Figures 3a–3c, but for negative NINO3 values. Areas within white contours are statistically significant at the 95% confidence level.

[13] The corresponding SST and outgoing longwave radiation (OLR) patterns are similarly obtained (Figures 3b, 3c, 3e, and 3f). The La Niña Modoki features warm anomalies in the far-eastern equatorial Pacific but strong cold anomalies near the Dateline (Figure 3b). The large negative anomaly appears to be rather effective in shifting convection westward, generating strong anomalies over the western tropical Pacific of the warm hemisphere, i.e., north of Australia and the eastern Indian Ocean (Figure 3c). This is why the impact of La Niña Modoki extends further to the west and has a significant impact over northwestern Australia. The associated southeastward moisture transport from the anomalous convective activities through northwest cloud-band events and interactions with regional synoptics lead to a rainfall increase into the northern Murray-Darling Basin region. By contrast, although La Niña cold anomalies occupy much of the equatorial Pacific, the maximum is situated in the eastern equatorial Pacific (Figure 3e). The impact of this pattern is the conventional-type response of a rainfall reduction mainly in the east, with corresponding OLR signals largely confined to eastern Australia (Figure 3f).

[14] There is a cyclonic wind circulation pattern associated with the westward-shifted low pressure centre, which forces a warm SST anomaly off the north western Australian coast (Figure 3b). However, we find that this is not the cause of the rainfall correlation, consistent with a previous finding that the oceanic anomaly there and rainfall anomalies over Australia are both driven by western Pacific SST anomalies [Watterson, 2001]. Indeed, the SST anomaly off the north western Australian coast lags the western Pacific SST anomaly by 2–3 months (statistically significant at the 95% confidence level), providing confirmation that it is a response to La Niña Modoki.

[15] A similar plot for El Niño and El Niño Modoki (Figure S1 of the auxiliary material) shows there are virtually no significant correlations with rainfall as expected. The SST pattern associated with NINO3 (Figure S1e) has a basin-wide Indian Ocean warm pattern conducive to autumn rainfall only over southern Western Australia (Figure S1d) with consistent OLR anomalies (Figure S1f). No equivalent signals are found for the El Niño Modoki (Figures S1aS1c).

6. Modulation on Conventional ENSO-Rainfall Teleconnection

[16] The modulation comes about through the asymmetry of the impacts in positive and negative phases of the indices, as already shown in Figures 2c and 2d. We now highlight the top five driest and wettest autumns from 1950–2007 (Figure 4, dots). Because there is virtually no relationship with rainfall in the El Niño phase, only one of the top five driest autumns occurred in conjunction with one of the top five highest NINO3 values (1993, Figure 4a). Likewise, only one of the top five driest autumns is in conjunction with one of the top five highest EMI values (1991, Figure 4b), the rest occur with neutral EMI values.

Figure 4.

Time series of (a) NINO3 and (b) EMI, with the top five wettest (blue) and driest (red) autumn rainfall seasons highlighted since 1950. The two horizontal lines indicate the one-standard deviation value of each index.

[17] Because of the significant influence of either type of La Niña, all the top 5 wettest autumns occurred when either the EMI or NINO3 was significantly negative, most with a very high negative EMI. In 1989, their superimposing effect contributed to the wettest autumn in terms of all-Australian rainfall since official records began in 1900, 108.8 mm above normal (based on the climatological mean over 1950–2007).

[18] An outstanding example of the combined impact from a non-relationship with El Niño and a strong relationship with La Niña Modoki is the autumn of 1983. In this autumn the value of NINO3 was at its highest positive value during the studied period, a residual from the 1982 event, leading NOT to a reduction in autumn rainfall, but a large positive anomaly (second wettest autumn) at 102.0 mm above normal. It is the strong influence from La Niña Modoki that plays a significant part in causing the anomalous autumn rainfall increase.

7. Conclusions

[19] We show that the impact from ENSO Modoki is similarly modulated by the IPO in all seasons except austral autumn (MAM), when its impact on Australian rainfall is markedly different from that of conventional ENSO. A focus on MAM finds that, in both types of ENSO, there is an asymmetry between the impacts from La Niña and from El Niño. During El Niño, no statistically significant relationship exists, whereas during La Niña, the relationship is significant. A cold anomaly associated with La Niña Modoki near the International Dateline is rather effective in shifting convection westward. As a result, a La Niña Modoki leads to a rainfall increase extending from northwestern Australia to the northern Murray-Darling Basin, whereas the influence of a conventional La Niña is mainly located to the east. The asymmetry, with a rainfall increase associated with a La Niña Modoki, provides a plausible explanation for the strong inter-ENSO variations in autumn Australia rainfall. This includes the autumn of 1983, when NINO3 recorded its highest positive value, yet it was the second wettest autumn since official records started in 1900. Our results highlight the difficulty in predicting the impacts on Australian autumn rainfall of either type of El Niño events, and underscore the importance of taking into account the influence of La Niña Modoki in predicting ENSO's impact.


[20] This work is supported by the Department of Climate Change, and CSIRO Weath from Oceans Flagship.