Projected multi-model mean changes between the historical period and the 21st Century under the RCP8.5 scenario, using all 24 models, are depicted in Fig. 2. The figure shows that the spatial structure of changes in the mean and in decile breaks 1–9 have a similar structure and a similar magnitude in most areas. This is consistent with external forcing driving a shift in the distribution of precipitation over the region to a lower value, with little change in the shape of the distribution. We will examine contrasts between the various panels, the corresponding shape changes, and their implications later.
Note again, that all of the results in this paper refer to Winter–Spring rainfall.
Changes linked to Decile 1
The value of the Decile 1 threshold declines over most of the country (Fig. 2, right hand upper panel), apart from the northwest, where there is little change. This panel is similar to Fig. 3, which shows the change in the percentage of years that are below the pre-industrial Decile 1 value, from the twentieth century to the twenty-first century. Values vary from a decrease of a few percent in the northwest, to increases of more 10% in most of southern Australia, with increases over 30% in the southwest and about 15–20% in Victoria. Approximately 60% of years are projected to fall below the preindustrial Decile 1 value near Perth in southwest WA, about 35% near Adelaide in South Australia, 30% of years near Melbourne, 19% of years near Sydney, 23% of years in Canberra, and about 25% of years near Brisbane.
Changes in the frequency of droughts occurring in consecutive Winter–Spring seasons
Here we examine droughts when consecutive Winter–Spring seasons all fall below the lowest decile (Decile 1) under preindustrial conditions. Consecutive years of low precipitation during the Winter–Spring season can be damaging to ecosystems and impact management of water, bushfires and agricultural systems. The number of “droughts” of length 1–6 consecutive years is depicted in Fig. 4 for the historical period and for the twenty-first century under the RCP8.5 scenario. Each NRM cluster is considered separately. The plots indicate that the multi-model median number of droughts of duration 1–4 years tends to increase in all of the NRM clusters in the twenty-first century relative to the twentieth century. During the twentieth century, 3-year droughts occur very infrequently. In fact, among the 24 simulations examined, two simulations at most produce three-year droughts in the NRM clusters. Four-year droughts are even less frequent, occurring in one simulation only or not at all in each NRM cluster.
In the twenty-first century under RCP8.5 the frequency of multi-model median one-year droughts increases in every cluster by between 1 and 5 events per century on average. While 5- and 6-year droughts do not occur during the twentieth century, they occur, albeit very infrequently, in the twenty-first century in all clusters. In all clusters there are 1–2 5- and six-year events, except in Southern and Southwestern Flatlands, where the frequency is higher on average.
The increase in the frequency of droughts occurring in consecutive Winter–Spring seasons is clearly an important component of the increase in the frequency of years in drought (i.e., regardless of duration) depicted in Fig. 3.
Changes in other deciles
If all of the maps in Fig. 2 were identical then this would suggest that external forcing causes precipitation (relative frequency) distributions for all regions to shift, with the shape of the distributions unchanged. However, the maps are not identical, indicating that in some regions external forcing may have also driven a change in the shape of the precipitation distribution. This effect is most clearly seen over the Solomon Islands and PNG near the northern and eastern edges of the plot. Here there is very little change in Decile 1 precipitation, whereas the changes in Deciles 6–9 are large. In this region Deciles 6–9 are higher in the twenty-first century than in the historical and pre-industrial periods.
Some of the differences between panels in Fig. 2 are presented in Fig. 5. The top left panel, in conjunction with Fig. 2, shows that the projected reduction in Decile 1 over much of the country is not as great as the projected reduction in Decile 5. There are also differences between the projected change in Decile 1 and Decile 9 precipitation (Fig. 5, bottom left). For example, Decile 9 increases markedly over PNG and the Solomon Islands whereas Decile 1 does not change much at all in the same region and the magnitude of the reduction in Decile 1 over Victoria and parts of southern NSW tends to be greater than the magnitude of the projected reduction in Decile 9 precipitation.
Asymmetries are also present in the projected changes for minima and maxima (Fig. 5, bottom right). This is also evident in Fig. 2, which shows that changes in the minima are similar to changes in most deciles, whereas maxima do not change much over most of the country, and actually increase in many locations. For example, over northwest Australia the minimum drops while the maximum monthly precipitation value increases, while over Victoria and part of eastern Australia the minimum declines, while the maximum value is unchanged.
In summary, in some locations external forcing, in addition to shifting precipitation distributions, also drives changes in the shape of the precipitation distribution. We will return to this issue below.
Combined impact of ENSO and external forcing
As noted in the model assessment section, we focus here on the five models that are best able to simulate ENSO teleconnections to Australian rainfall.
The proportion of El Niño years, La Niña years and neutral years resulting in Decile 1 precipitation in the five models with the most realistic ENSO teleconnections is depicted in Fig. 6 during the twentieth century (left column) and the twenty-first century (middle column). The difference between them (i.e., RCP8.5-historical) is also presented (right column).
During the historical period over eastern Australia approximately 15–30% of El Niño years result in Decile 1 precipitation, whereas only around 4% of La Niña years do. These figures are much larger during the twenty-first century: approximately 25–40% of El Niño years result in Decile 1 precipitation, while approximately 12% of La Niña years do.
In southwest WA approximately 20% of El Niño years and 12% of La Niña years result in Decile 1 precipitation during the twentieth century, whereas during the twenty-first century these figures increase to 70% and over respectively.
The combined impact of ENSO and climate change on precipitation in the eight NRM clusters is depicted in Fig. 7, which shows standardized precipitation (see caption) distributions for El Niño, La Niña and all years, for the twentieth and twenty-first centuries. The plot indicates that ENSO influences all eight clusters, that El Niño years tend to be drier than La Niña years, and climate change does not change this situation. However, both El Niño and La Niña events tend to be drier during the twenty-first century than they were in the twentieth century. This is as expected from the previous discussion: the twenty-first century distributions for all years, El Niño, and La Niña tend to shift to lower values relative to their twentieth century counterparts. However, we saw above that the results could not be fully explained in terms of shifts alone. The same is true in seven of the eight panels in Fig. 7 (excluding the Southern and South-western Flatlands): the extreme right-hand (wettest) part of the distributions tend to be little changed or less changed than are the extreme left (driest) parts of the distributions. In fact, in these seven clusters the distributions for La Niña years actually extend further to the right during the twenty-first century than their twentieth century counterparts, albeit marginally. This is consistent with the contrast in the changes in minima and maxima described above (Figs. 2, 5).
Changes in ENSO variability
We examined changes in ENSO-driven variability—measured using E–L, the difference between El Niño and La Niña precipitation composites—in the eight NRM clusters (Fig. 1) from 1940–1989 to 2050–2099. We found that a while a majority of the five models best able to simulate ENSO teleconnections showed an increase in |E–L| in all clusters, the majority was very marginal in all clusters except the Wet Tropics (4 out of 5 models showing an increase), the Monsoonal North (4 out of 5), the Central Slopes (4 out of 5) and Southern and Southwestern Flatlands (4 out of 5). Increases in precipitation during La Niña years are projected to occur in the Wet Tropics and Monsoonal North (4 out of 5), while decreases in precipitation during La Niña years are projected for the Southern Slopes (4 out of 5) and Southern and Southwestern Flatlands (5 out of 5). Precipitation during El Niño years is projected to decrease in the Wet Tropics (4 out of 5), the Murray Basin (5 out of 5), the Southern Slopes (5 out of 5), the East Coast (4 out of 5), the Rangelands (4 out of 5) and Southern and Southwestern Flatlands (5 out of 5).
We also examined changes in precipitation variability between the twentieth and twenty-first centuries. This includes, but is not restricted to, variability driven by ENSO. The results were very marginal. While the multi-model mean variability increased in six of the eight clusters, there is a lot of model-to-model differences and not all models exhibit an increase. The most robust increases occurred in the Southern Slopes where all five models that are best able to simulate ENSO teleconnections project an increase, while four out of the same five models project an increase in the Southern and Southwestern Flatlands.