1. Introduction
Over recent decades severe rainfall reductions have occurred in regions such as southwest Western Australia (e.g., Cai and Cowan 2006; Ummenhofer et al. 2008; Hope et al. 2009) and southeast Australia (e.g., Cai and Cowan 2008; Nicholls 2009). In contrast, austral summer [December–February (DJF)] rainfall over northwest Australia (NWA) has increased by about 0.8% yr−1 since 1950. Summer is the major rainfall season for the region, accounting for more than half of the annual total. If this wet season trend continues, it could become an important future water resource for northern regional communities.
A point of contention is whether increasing anthropogenic aerosols have played a part in the observed rainfall increase. Comparing two sets of experiments with and without increasing anthropogenic aerosols, using the CSIRO Mk3A model, Rotstayn et al. (2007) suggest a possible impact from increasing Asian aerosols; this impact is achieved through an alteration of the north–south surface temperature and pressure gradients in the tropical Indian Ocean. Subsequent warming in the eastern Indian Ocean induces an enhanced northwesterly monsoonal flow toward NWA. However, Rotstayn et al. (2007) describe reservations with the model that include its inability to correctly simulate interdecadal variability in Australian rainfall and the lack of model agreement with observations over eastern Australia.
As the majority of anthropogenic aerosols are projected to decrease by the end of the twenty-first century, including parts of Asia (Streets 2007), the inference is that, if aerosols control the observed NWA rainfall trend, then long-term future rainfall over NWA will decrease. This matches projections of a future rainfall decrease over NWA by the Fourth Assessment Report (AR4) of the Intergovernmental Panel on Climate Change (IPCC), although there is significant variation in the magnitude of change among models contributing to the all-model ensemble (CSIRO 2007; Meehl et al. 2007). Shi et al. (2008) show that the modeled impact of aerosols on NWA rainfall in the CSIRO Mk3A is a consequence of the model generating an Indian Ocean dipole (IOD) in DJF. The model also trends toward a negative IOD-like state [anomalously warm (cold) SSTs in the eastern (western) Indian Ocean] in DJF since 1950, which results in an increasing rainfall trend in NWA in this season, as seen in observations. In reality, the IOD, which describes the zonal fluctuations of sea surface temperature (SST) and winds in the Indian Ocean, emerges in austral winter and terminates in late austral spring (Saji et al. 1999; Ashok et al. 2003). No IOD can survive into summer due to the onset of intraseasonal disturbances and the reversal of monsoonal winds (Rao and Yamagata 2004). However, Luffman et al. (2010) show that forcing an atmospheric model with historical tropical Indian Ocean SST trends produces a drying trend across much of northern Australia, with a small rainfall increase in central Australia. In their model, the Indian Ocean surface warming induces upper-level (surface) divergence (convergence) driven by strong convective uplift, leading to the suppression of convection in the surrounding regions including NWA. Other studies (Taschetto and England 2008; Zhang 2009; Berry et al. 2011) suggest the early onset and increased duration of the Australia–Asia summer monsoon may be responsible for the NWA rainfall increase. Similar observations of a prolonged monsoon have been seen in parts of central Asia (Zhang 2009). Adding to the complexity of the monsoon’s influence on NWA rainfall is the recently discovered El Niño–Southern Oscillation (ENSO) Modoki (Ashok et al. 2007; Wang and Hendon 2007); El Niño Modoki events have been shown to intensify NWA rainfall, particularly in January and February (Taschetto et al. 2010).
In this paper, we use outputs of the twentieth-century experiments from 24 models made available as part of phase 3 of the Coupled Model Intercomparison Project (CMIP3) to assess the level of consensus in terms of the impact from increasing anthropogenic aerosols. An advantage of an ensemble mean is that, when aggregated over multiple models and multiple experiments, a contribution to a trend by variability is largely cancelled out because each model and experiment has its own variability, leaving mostly the response to climate change. Further, consequences arising from biases of particular models may be offset or diluted. As we will show, results from CMIP3 models suggest that aerosols play an insignificant role. We then proceed to examine the implication for future rainfall projections using twenty-first-century climate experiments.
2. Model experiments, reanalysis, and observations
To examine if climate models produce the observed DJF rainfall trend, we analyze 50 years (1950–99) from multiple twentieth-century experiments using 24 CMIP3 models. Model names and origins are listed in Table 1. Many models have multiple ensemble members (see Fig. 2), providing a total of 75 experiments. Here the ensemble mean of each model is presented. All CMIP3 models include the direct effect of anthropogenic aerosols, but only 10 models incorporate the additional indirect effect (Fig. 2 details which models include the indirect effect). Since the single-model experiments in the CSIRO Mk3A (Rotstayn et al. 2007) contain both aerosol effects, to facilitate a better comparison we stratify the experiments into two groups, those with and those without the indirect effect of aerosols. We also use a multicentury-long control experiment (without any climate change forcing) from the CSIRO Mk3A to assess the impact of multidecadal variability.
Acronym and name (source) of models used in ensemble experiments.
Linear trends of DJF rainfall are calculated for Australia over the 1950–99 period for each of the 24 CMIP3 models, as well as for the all-model average (75 experiments), the 14-model average (43 experiments) with the direct aerosol effect only, and the 10-model (32 experiments) average with both the direct and indirect effects. The changes are presented in percentage of climatology to take into account that the climatological value over NWA could be rather different in individual models.
To examine the process responsible for future rainfall projections across NWA, we use outputs from twenty-first-century experiments for the period 2001–2100 to examine the future rainfall change, expressed in terms of percentage change per degree Celsius of global warming. Changes in mean sea level pressure are scaled in a similar way. This way of scaling allows a comparison between different emission scenario projections (Mitchell 2003) and was adopted by the IPCC for the AR4. We explore the potential link between future rainfall changes and the present-day simulation of mean climate and teleconnection with climate drivers.
Also deployed are reanalyses from the National Centers for Environmental Prediction (Kalnay et al. 1996), observed Australian rainfall since 1900 from the Bureau of Meteorology Research Centre (Lavery et al. 1997; Jones et al. 2009), and reconstructed SST from the Met Office Hadley Centre (Rayner et al. 2003).
3. Rainfall trends from twentieth-century experiments
Figure 1 shows Australian summer rainfall trend maps from the observed and ensemble means. Several results emerge, the first of which is that none of the CMIP3 model ensemble means (Figs. 1b–d) produces the observed rainfall trend pattern of an increase in western Australia and a decrease in eastern Australia (Fig. 1a). Second, although the simulation with the indirect aerosol effect (Fig. 1c) is slightly closer matched to the observed than the case without indirect aerosols (Fig. 1d), neither case adequately captures the observed changes, suggesting that the indirect effect has little impact. Although each ensemble produces a tendency of a weak increase in rainfall over the western half, the trend over eastern Australia is somewhat opposite to the observed. We also examine the patterns in each individual model; only 4 out of 24 models [CSIRO Mk3.5, GFDL CM2.1, MIROC3.2 (hires), and UKMO HadCM3] generate a pattern that is broadly similar to the observed (not shown).
Map of Australian DJF rainfall trends from (a) observations (1950–2008), (b) an ensemble mean of 24 CMIP3 models, and an ensemble mean of CMIP3 models (c) with and (d) without the indirect effect of anthropogenic aerosols. All model ensemble trends are from 1950 to 1999 and are expressed as a climatological change per summer (percentage per summer). An outline of NWA is shown in (a).
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3832.1
Focusing on trends over NWA (averaged over 10°–25°S, 110°–135°E; land points only, see Fig. 1 for outline), the results are summarized in Fig. 2. The uncertainty range of the trend is estimated as the standard error of the linear regression fit on the ensemble-mean data, taking into account the number of ensemble members of each model. Only nine models produce a statistically significant increasing rainfall trend, although all are much weaker than the observed (1950–2008), which is 0.8% yr−1 (summer) or nearly 50% of the climatological mean since 1950. The largest model trend is generated by GFDL CM2.0, which does not incorporate the indirect aerosol effect, with a trend less than half of the observed. The all-model ensemble mean trend is only 0.05% yr−1, while the ensemble that incorporates both the direct and indirect aerosol effects shows a slightly greater trend at 0.08% yr−1. Of the nine models that produce an increase, six are forced with only the direct effect of aerosols (Fig. 2, red circles) and three are forced with the both aerosol effects (Fig. 2, blue circle).
DJF rainfall trends in terms of climatological change (% yr−1) from observations (1950–2008, blue triangle), and CMIP3 twentieth-century climate model experiments [1950–99, blue (red) circles indicate models with (without) the indirect effect of anthropogenic aerosols]. Error bars indicate the standard error of the linear regression. The number of individual members in each ensemble is shown in parentheses next to the model names.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3832.1
The multimodel ensemble (24 models with a total of 75 experiments) should be more reliable than using any particular individual model, as the aggregation effectively removes any possible impacts from multidecadal variability and dilutes the impact of any biases from individual models. However, the results indicate that increasing anthropogenic aerosols in the multimodel ensemble play a negligible role in generating the observed NWA rainfall increase. More importantly, because those models that contain increasing aerosols do not replicate the observed increase, it follows that the projected rainfall reduction is not due to a projected decrease in aerosols. Questions then arise as to what processes control the future rainfall decrease as projected by the IPCC AR4 and whether the processes are realistic. These questions are addressed in section 4.
Given that the CMIP3 model ensemble result is vastly different from the result of the CSIRO Mk3A, which actually produces the observed increase (Rotstayn et al. 2007), we provide a brief focus on this model. We examine outputs of the multicentury preindustrial control experiment from the CSIRO Mk3A to examine the possible role of natural multidecadal variability. A time series of trends over multiple 50-yr periods is constructed using a sliding window, with each point representing the trend value centered over that year. Over a 50-yr period, positive trends comparable to the observed in terms of total rainfall amount and percentage of climatology (Fig. 3a) are entirely possible without a climate change forcing. However, this result does not insinuate that the model produces the correct physics of rainfall variability. In this model, the eastern pole of the IOD in DJF is a part of the western Pacific warm pool (Shi et al. 2008), which varies with ENSO as a consequence of a severe westerly bias of the Pacific cold tongue (Cai et al. 2003).
(a) Time series of 50-yr trends of NWA DJF rainfall calculated using a 50-yr sliding window on outputs of a multicentury-long control experiment from the CSIRO Mk3A without climate change forcing. The trend from the observations (1950–2008, blue triangle) is shown on the right side. (b) Observed eastern Indian Ocean DJF SST anomalies (averaged over 5°–15°S, 90°–110°E) against NWA rainfall. (c) As in (b), but based on the 300-yr output using the control experiment from the CSIRO Mk3A.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3832.1
This spurious IOD in DJF in the CSIRO Mk3A produces a rainfall teleconnection that is opposite to the observed and contributes to the result that aerosols could force a rainfall increase across NWA. During El Niño, the observed DJF rainfall over NWA tends to decrease, with basinwide warming over the Indian Ocean (including the eastern Indian Ocean). This is a consequence of a well-known atmospheric teleconnection in which a decreased Walker circulation leads to easterly anomalies (Klein et al. 1999; Alexander et al. 2002). As such, a positive eastern Indian Ocean SST anomaly (averaged over 5°–15°S, 90°–110°E) is weakly associated with a decrease in rainfall (Fig. 3b). In the model, the spurious IOD produces the relationship that is directly opposite (Fig. 3c): anomalously cool eastern Indian Ocean SSTs are associated with anomalously low rainfall over NWA. Because of the hemispheric asymmetry of the geographic distribution of anthropogenic aerosols (the major source is the Northern Hemisphere), a greater surface cooling is generated over the northern Indian Ocean sector than over the southern Indian Ocean (Cai et al. 2007). As a combined effect of this thermal contrast and the Coriolis force, trends of northwesterly anomalies, reminiscent of those during a negative IOD, are generated over the southern tropical Indian Ocean, particularly along the Sumatra–Java coast. These wind anomalies lead to a larger warming over the eastern Indian Ocean than over the western Indian Ocean. Following the relationship shown in Fig. 3c, a rainfall increase is generated over NWA. Thus, the rainfall increase in the CSIRO Mk3A is generated in association with the DJF IOD and its rainfall teleconnection.
4. A contributing factor for the projected rainfall decrease
What then contributes to the projected rainfall decrease? It is recognized that anthropogenic climate change signals project onto existing natural variability modes (Stone et al. 2001), such as the southern annular mode (Marshall 2003) and the IOD (Cai et al. 2009a), leading to enhanced impacts on the climate (e.g., rainfall) over many regions of Australia. Thus, it is appropriate to examine whether the degree of realism in the simulation of the associated rainfall teleconnection has contributed to the projected rainfall decrease in summer.
To this end, using outputs from one experiment of each model, we examine the NWA summer rainfall teleconnection with an important climate mode impacting this region, ENSO. We regress gridpoint detrended rainfall anomalies onto a detrended ENSO index represented by SST anomalies averaged over the Niño-3.4 region (5°S–5°N, 170°–120°W). To gauge the significance of the ENSO–NWA rainfall teleconnection, we correlate gridpoint detrended rainfall anomalies with the ENSO index. The maps of regression and correlation coefficients are then averaged. Compared with the observed, the largest correlations in the all-model average are recorded over NWA (Fig. 4a). These correlations are significant at the 95% confidence level, for which a 50-yr sample size requires a correlation coefficient greater than 0.28. In reality, the ENSO influence on rainfall predominantly occurs over northeast Australia rather than NWA only (Fig. 4b). The models perform reasonably well at capturing the rainfall teleconnection over western NWA; however, they struggle to simulate ENSO’s control on rainfall over northeastern Australia. This means that during El Niño, the simulated rainfall over NWA decreases instead of the concurrent stronger reductions expected over northeastern Australia.
Variations of DJF rainfall associated with ENSO based on (a) the 24-model average over the Indo-Pacific region and (b) observations over Australia. Variations describe detrended gridpoint DJF rainfall anomalies regressed onto a detrended ENSO index of SSTs averaged over the Niño-3.4 region (mm °C−1). Black contours indicate statistically significant correlations at the 95% confidence level.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3832.1
Is the strength of the ENSO–NWA rainfall teleconnection relevant to the future NWA rainfall changes? It turns out to be highly relevant, as illustrated by a scatterplot of intermodel variations of NWA DJF rainfall changes for the 2001–2100 period versus intermodel variations of the ENSO–NWA rainfall teleconnection (Fig. 5). We include only 23 models because the PCM1 future rainfall change is not available. The teleconnection over NWA for each model (the observations are represented by a blue line) confirms that many of the models produce an overly strong correlation. Further, there is a tendency for models with a greater teleconnection to generate a greater future NWA rainfall reduction. Such a tendency, as measured by a linear fit, is statistically significant at the 95% confidence level (correlation greater than 0.41). What this result means is that a realistic simulation of the present-day rainfall teleconnection has a bearing on future rainfall projections for the region.
Intermodel variations of NWA rainfall changes from 2001 to 2100 vs intermodel variations of the present-day ENSO–NWA rainfall teleconnection, defined as the correlation between NWA rainfall and an ENSO index in DJF for 1950–99 (both detrended). This relationship is based on 23 models in which outputs for the future rainfall change are available. The rainfall trends are expressed as a percentage change per degree of global warming per century [% °C−1 (100 yr)−1] to take into account the differences in model rainfall climatology and in different emission scenarios. The vertical blue line shows the observed ENSO–NWA rainfall teleconnection.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3832.1
To confirm this, we construct a gridpoint rainfall correlation with ENSO C(x, y, k) and ENSO–rainfall signal-to-noise ratio R(x, y, k), with k representing each of the 23 models. The field of C(x, y, k) is obtained by correlating Niño-3.4 with gridpoint rainfall anomalies. The ENSO rainfall “signal” is defined as the standard deviation of rainfall anomalies associated with ENSO and is determined from a linear regression onto Niño-3.4. “Noise” is defined as the standard deviation of the residual after removing ENSO-related rainfall signals. As expected, a large ratio indicates that the ENSO signal is able to manifest from the background noise; therefore, the greater the ratio, the stronger the ENSO–rainfall teleconnection. We then correlate the future rainfall change FRC(x, y, k) (again k representing each of the 23 model) with C(x, y, k) and R(x, y, k) with respect to k. This is the same as calculating the correlation coefficient of the linear fit shown in Fig. 5 except that this is carried out over all grid points.
The result, shown in Fig. 6, highlights several important features. First, the influence of the ENSO–rainfall teleconnection, or the ENSO rainfall signal-to-noise ratio, on the rainfall projection is strongest over the eastern Indian Ocean and the western Pacific encompassing NWA. The well-defined pattern underpins the systematic nature of the influence by the rainfall teleconnection or the signal-to-noise ratio. Second, the stronger the ENSO–rainfall teleconnection or the ENSO signal-to-noise ratio, the greater the future rainfall reduction, consistent with Fig. 5.
Correlation of gridpoint future rainfall changes over the 2001–2100 period with (a) present-day ENSO–rainfall teleconnection and (b) present-day ENSO rainfall signal-to-noise ratio. See section 4 for more details. Contours indicate statistically significant correlations at the 95% confidence level (an absolute value greater than 0.41).
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3832.1
Because the ENSO–NWA rainfall teleconnection is unrealistically strong, it follows that the projected rainfall reduction is also unrealistically strong. What drives the unrealistic teleconnection? An important contributing factor is the well-known model Pacific cold tongue and warm pool, which extends too far west into the western Pacific and eastern Indian Ocean (e.g., Davey et al. 2002). As a consequence of this bias, the ascending branch of the Walker circulation also extends too far west (Cai et al. 2009b). As the Walker circulation oscillates with the model’s ENSO cycle, that is, shifts eastward and weakens during El Niño, an ENSO–NWA rainfall relationship is generated (e.g., see Fig. 3 of Cai et al. (2009b)) in addition to the unrealistic climatological rainfall (Lin 2007). This is rather problematic because, in reality, the largest teleconnection operates over northeast Australia (Fig. 4b).
Why do models with a greater ENSO–NWA rainfall teleconnection produce a greater future rainfall reduction across NWA? Previous studies have shown that the Walker circulation weakens in a warming climate (e.g., Vecchi et al. 2006). As the global warming signals tend to project onto existing modes of natural variability (Stone et al. 2001), one expects that models with a greater ENSO amplitude will also show a greater weakening in the Walker circulation. This is indeed the case as shown in Fig. 7. We note that, although the majority of models do produce a weakening of the Walker circulation, some models do not; for example, there are four models that show a considerable strengthening of the Walker circulation (GISS-AOM, GISS-ER, INGV ECHAM4, ECHO-G), that is, a La Niña–like state. If we regard these four models as outliers and exclude them, the models that simulate a greater weakening in the Walker circulation tend to also produce a greater future rainfall reduction over NWA (figure not shown) where the model Walker circulation has an impact. This relationship should predominantly occur over northeast Australia rather than NWA if the models simulate the observed ENSO–rainfall teleconnection.
Intermodel variations from 23 models of Walker circulation changes from 2001 to 2100 vs the present-day ENSO amplitude, defined as the standard deviation of each model’s Niño-3.4 index for 1950–99. The Walker circulation is defined as the difference in sea level pressure anomalies between the west Pacific (5°S–5°N, 80°–160°E) and east Pacific (5°S–5°N, 160°–80°W), as described in Vecchi et al. (2006). The Walker circulation trend is expressed as a surface pressure change per degree of global warming per century [hPa °C−1 (100 yr)−1].
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3832.1
5. Conclusions
The hypothesis that the observed increasing summer rainfall trend over NWA is a result of increasing anthropogenic aerosols is one that has yet to be conclusively tested. This study uses outputs of 24 models with a total of 75 experiments to provide a multimodel perspective. We find that none of the CMIP3 ensemble averages of experiments with both the direct and indirect anthropogenic aerosol effects included (10 models), or only the direct aerosol effect (14 models), simulate the observed rainfall increase over NWA. Given this, it follows that the projected rainfall reduction in these models is not due to a projected decline in future aerosol emissions. We then proceed to address what processes may contribute to the projected rainfall decrease. We find that the projected reduction is at least in part associated with an unrealistic rainfall teleconnection that is too strong. Previous studies have shown that this unrealistic feature arises from a common model Pacific cold tongue/warm pool bias. As a consequence of the bias, the warm pool, the mean convection center, and the associated ascending branch of the Walker circulation are situated too far west. As such their longitudinal fluctuations with the model ENSO cycle generate a rainfall teleconnection that is greatest over NWA rather than northeast Australia. Under global warming, the weakening in the Walker circulation is generally greater in models with a greater ENSO amplitude, as some global warming signals project onto ENSO. It is these models with a greater ENSO amplitude, a greater NWA rainfall teleconnection, and a future weakening of the Walker circulation that also produce a greater future NWA rainfall reduction. Our study suggests that a realistic simulation of the ENSO–NWA rainfall teleconnection is essential for reducing uncertainty of rainfall projections for the region and that the Pacific cold tongue/warm pool bias is one such uncertainty source that needs to be alleviated.
Acknowledgments
This work is supported by the Australian Department of Climate Change and Energy Efficiency. We recognize the many modeling groups that have submitted output to the Program for Climate Model Diagnosis and Intercomparison that has produced the CMIP3 database. We also thank Peter van Rensch for his assistance in producing some of the figures.
REFERENCES
Alexander, M. A., I. Bladé, M. Newman, J. R. Lazante, N.-C. Lau, and J. D. Scott, 2002: The atmospheric bridge: The influence of ENSO teleconnections on air–sea interaction over the global oceans. J. Climate, 15, 2205–2231.
Ashok, K., Z. Guan, and T. Yamagata, 2003: Influence of the Indian Ocean dipole on the Australian winter rainfall. Geophys. Res. Lett., 30, 1821, doi:10.1029/2003GL017926.
Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata, 2007: El Niño Modoki and its possible teleconnection. J. Geophys. Res., 112, C11007, doi:10.1029/2006JC003798.
Berry, G., M. Reeder, and C. Jakob, 2011: Physical mechanisms regulating summertime rainfall over northwestern Australia. J. Climate, in press.
Cai, W., and T. Cowan, 2006: SAM and regional rainfall in IPCC AR4 models: Can anthropogenic forcing account for southwest Western Australian winter rainfall reduction? Geophys. Res. Lett., 33, L24708, doi:10.1029/2006GL028037.
Cai, W., and T. Cowan, 2008: Dynamics of late autumn rainfall reduction over southeastern Australia. Geophys. Res. Lett., 35, L09708, doi:10.1029/2008GL033727.
Cai, W., M. A. Collier, H. B. Gordon, and L. J. Waterman, 2003: Strong ENSO variability and a super-ENSO pair in the CSIRO Mark 3 coupled climate model. Mon. Wea. Rev., 131, 1189–1210.
Cai, W., T. Cowan, M. Dix, L. Rotstayn, J. Ribbe, G. Shi, and S. Wijffels, 2007: Anthropogenic aerosol forcing and the structure of temperature trends in the southern Indian Ocean. Geophys. Res. Lett., 34, L14611, doi:10.1029/2007GL030380.
Cai, W., A. Sullivan, and T. Cowan, 2009a: Climate change contributes to more frequent consecutive positive Indian Ocean dipole events. Geophys. Res. Lett., 36, L23704, doi:10.1029/2009GL040163.
Cai, W., A. Sullivan, and T. Cowan, 2009b: Rainfall teleconnections with Indo-Pacific variability in the WCRP CMIP3 models. J. Climate, 22, 5046–5071.
CSIRO, 2007: Climate change in Australia: Technical report 2007. CSIRO and Bureau of Meteorology Tech. Rep., 148 pp.
Davey, M. K., and Coauthors, 2002: STOIC: A study of coupled model climatology and variability in tropical ocean regions. Climate Dyn., 18, 403–420.
Hope, P., B. Timbal, and R. Fawcett, 2009: Associations between rainfall variability in the southwest and southeast of Australia and their evolution through time. Int. J. Climatol., 30, 1360–1371, doi:10.1002/joc.1964.
Jones, D., W. Wang, and R. Fawcett, 2009: High-quality spatial climate data-sets for Australia. Aust. Meteor. Oceanogr. J., 58, 233–248.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Klein, S. A., B. J. Soden, and N.-C. Lau, 1999: Evidence for a tropical atmospheric bridge. J. Climate, 12, 917–932.
Lavery, B., G. Joung, and N. Nicholls, 1997: An extended high-quality historical rainfall dataset for Australia. Aust. Meteor. Mag., 46, 27–38.
Lin, J.-L., 2007: The double-ITCZ problem in IPCC AR4 coupled GCMs: Ocean–atmosphere feedback analysis. J. Climate, 20, 4497–4525.
Luffman, J. J., A. S. Taschetto, and M. H. England, 2010: Global and regional climate response to late twentieth-century warming over the Indian Ocean. J. Climate, 23, 1660–1674.
Marshall, G. J., 2003: Trends in the southern annular mode from observations and reanalyses. J. Climate, 16, 4134–4143.
Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–845.
Mitchell, T. D., 2003: Pattern scaling: An examination of the accuracy of the technique for describing future climates. Climatic Change, 60, 217–242.
Nicholls, N., 2009: Local and remote causes of the southern Australian autumn-winter rainfall decline, 1958–2007. Climate Dyn., 34, 835–845, doi:10.1007/s00382-009-0527-6.
Rao, S. A., and T. Yamagata, 2004: Abrupt termination of Indian Ocean dipole events in response to intraseasonal disturbances. Geophys. Res. Lett., 31, L19306, doi:10.1029/2004GL020842.
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, 4407, doi:10.1029/2002JD002670.
Rotstayn, L. D., and Coauthors, 2007: Have Australian rainfall and cloudiness increased due to the remote effects of Asian anthropogenic aerosols? J. Geophys. Res., 112, D09202, doi:10.1029/2006JD007712.
Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360–363.
Shi, G., W. Cai, T. Cowan, J. Ribbe, L. Rotstayn, and M. Dix, 2008: Variability and trend of north west Australia rainfall: Observations and coupled climate modeling. J. Climate, 21, 2938–2959.
Stone, D., A. J. Weaver, and R. J. Stouffer, 2001: Projection of climate change onto modes of atmospheric variability. J. Climate, 14, 3551–3565.
Streets, D. G., 2007: Dissecting future aerosol emissions: Warming tendencies and mitigation opportunities. Climatic Change, 81, 313–330.
Taschetto, A. S., and M. H. England, 2008: An analysis of late 20th century trends in Australian rainfall. Int. J. Climatol., 29, 791–807.
Taschetto, A. S., R. J. Haarsma, A. Sen Gupta, C. C. Ummenhofer, K. J. Hill, and M. H. England, 2010: Australian monsoon variability driven by a Gill–Matsuno-type response to central west Pacific warming. J. Climate, 23, 4717–4736.
Ummenhofer, C. C., A. Sen Gupta, M. J. Pook, and M. H. England, 2008: Anomalous rainfall over southwest Western Australia forced by Indian Ocean sea surface temperatures. J. Climate, 21, 5113–5134.
Vecchi, G. A., B. J. Soden, A. T. Wittenberg, I. M. Held, A. Leetmaa, and M. J. Harrison, 2006: Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature, 441, 73–76, doi:10.1038/nature04744.
Wang, G., and H. H. Hendon, 2007: Sensitivity of Australian rainfall to inter–El Niño variations. J. Climate, 20, 4211–4226.
Zhang, H., 2009: Diagnosing Australia-Asian monsoon onset/retreat using large-scale wind and moisture indices. Climate Dyn., 35, 601–618, doi:10.1007/s00382-009-0620-x.
