Synoptic components of rainfall variability and trends in southeast Australia

Authors


ABSTRACT

This work examines the contributions of different synoptic types to rainfall variability and trends in southeast Australia, with a focus on the drought of 1996–2009. Rainfall in the Mallee region is used to characterize the southeast region. The reduction in rainfall in the Mallee region during the drought is about two-thirds composed of a reduction in rainfall from cutoff lows and about one-third due to a reduction in rainfall from frontal systems. The reduction in frontal rain is mostly due to a reduction in rain per frontal system, which is associated with a reduction of baroclinicity in the southeast and south of Australia. The reduction of cutoff rainfall is mostly due to a reduction in number of the most intense cutoff systems. The frequency of cutoff systems matches changes in blocking activity in the Tasman Sea region. Blocking has undergone a weak decline in the Tasman region over the period of the drought. Analysis of synoptic system contributions to the drought indicates that causal explanations of the drought should account for variation in cutoff systems and blocking.

1. Introduction

Southeastern Australia was in a prolonged drought between 1996 and 2009 (Fawcett, 2004; Gallant et al., 2007; Watkins and Trewin, 2007; Murphy and Timbal, 2008; Timbal, 2009; Watterson, 2010). This recent drought was marked by an absence of high rainfall years (Pook et al., 2009; Ummenhofer et al., 2009) and stands out as one of the most significant droughts in the instrumental record covering the past century.

A variety of different causes and processes have been proposed as the primary or substantial source of the drought. These include association with the El Niño Southern Oscillation (ENSO) (Gallant et al., 2007), the Indian Ocean Dipole (IOD) (Cai et al., 2009; Ummenhofer et al., 2009), Southern Annular Mode (SAM) (Meneghini et al., 2007; Verdon-Kidd and Kiem, 2009; Nicholls, 2010b), subtropical ridge (Timbal, 2009; Nicholls, 2010a; Timbal and Hendon, 2011), and storm track growth modes (Frederiksen and Frederiksen, 2011). To be sure, most of these studies recognize the importance of a range of different driving processes (Risbey et al., 2009b) in determining the variability of rainfall. Some of these same processes have also been ruled out as possible causes of the drought. For example, some conclude that the drought could not be associated with ENSO (Murphy and Timbal, 2008; Nicholls, 2010b; Timbal and Hendon, 2011), not with the IOD (Verdon-Kidd and Kiem, 2009; Nicholls, 2010b; Timbal and Hendon, 2011), and not with SAM (Timbal, 2009). The existing literature is therefore somewhat contradictory, with some studies ruling out some of the processes that other studies have found to be a substantial cause.

Most of these conclusions about which processes are driving rainfall trends in southeast Australia are based on statistical associations between rainfall in the region and an index of the relevant process. These compare the timing and pattern of the rainfall changes with that expected from different driving processes to help establish the cause of trends. For example, the rainfall decline in the region is particularly pronounced in autumn (Murphy and Timbal, 2008), which is consistent with trends in the subtropical ridge at that time of year, which shows an intensification appropriate to reduce rainfall (Timbal, 2009).

In this work, we take a different approach based on synoptic types. We use a synoptic classification of weather systems in southeast Australia, relating each rainfall event to the synoptic system that produced it. That allows us to track the variation in rainfall according to variation in the number of synoptic systems of different type, which may in turn have implications for causes of the trends. The approach takes a rainfall trend as starting point and asks:

  • Are there clear contributions to the trend from different types of synoptic systems?

  • Do these contributions change as a result of changes in the frequency, intensity, or other aspects of these systems?

  • What do these contributions imply about any changes in regional atmospheric circulation?

  • Can changes in regional circulation be linked to other factors or broadscale circulation changes?

  • What does all this imply, if anything, about causes of the rainfall trends?

In the following sections, we outline the data and methods used, and the rainfall trends derived for each synoptic type. Following that we examine changes in circulation that are consistent or inconsistent with the synoptic components of the rainfall trends.

2. Methods and data

For the purposes of this paper, the southeastern Australian region is roughly the region south of 30°S and east of 135°E on the Australian continent. Mean annual rainfall in southeastern Australia varies from a few hundred millimetres in parts of the continental interior of this region to over a metre in the mountains of the Great Dividing Range and western Tasmania. A range of different synoptic systems produce rainfall across the southeast, dominated by cutoff low systems, frontal systems, and stream flow in winter and the transition seasons (Pook et al., 2006). Summer convective rainfall can be important locally, but is generally a smaller contributor to annual totals in the southeast. We focus here on winter and the transition seasons when rainfall is more coherently associated with distinct synoptic systems.

A given circulation or synoptic system may have very different impacts on rainfall across southeastern Australia. For example, a frontal system may bring rains to the southern, but not the northern part of the region, or a southeasterly stream may produce good rainfall in the areas between the Great Dividing Range and the coast in the very southeast, but little elsewhere. Because rainfall varies across the region for any given system, we do not use a single southeast region index of precipitation. Our attempt here is to relate rainfall to the synoptic systems that produced it, and as such it is preferable that the region we choose responds fairly uniformly to a given synoptic system. Thus, we have selected the Mallee region within southeast Australia for analysis of rainfall (Figure 1). This subregion is well removed from the major topographic features in the southeast, and rainfall is relatively homogeneous across the region. Further, rainfall in the Mallee region is highly correlated with rainfall across the southeast (Figure 2) and is broadly indicative of trends in the region (Brown et al., 2009).

Figure 1.

Map of Australia showing the locations of the synoptic analysis box and the Mallee rainfall stations. The synoptic box spans latitudes 30–45°S and longitudes 125–147.5°E. The background field in contours is a climatology of April–October rainfall for the period 1900–2009. All correlations shown are positive.

Figure 2.

Correlation map of April–October rainfall for the Mallee eight stations with gridded rainfall from the AWAP (Jones et al., 2009) dataset. Only correlations significant at the 95% level are shown. The years for this analysis are 1948–2007. All correlations shown are positive.

There are some regions evident in Figure 2 where rainfall is not well correlated with Mallee rainfall. These regions are the southeast coast and western Tasmania, where interactions between topography and stream conditions are important. The results presented here would not apply to these subregions in southeastern Australia. These regions would require a dedicated analysis to cater for the differences in synoptic climatology there.

Rainfall in the Mallee region is represented here by the average of rain over eight stations in the region (plotted in Figure 1). A description of the rainfall stations and the characteristics of rainfall in the region are given in Risbey et al. (2009a). The stations are all part of the Bureau of Meteorology high quality Australian historical dataset (Lavery et al., 1997). Any gaps in the Bureau of Meteorology records have been filled with interpolated data as documented in Jeffrey et al. (2001). We also used gridded rainfall data to represent rainfall changes across the continent. The gridded rainfall dataset is the Australian Water Availability Project (AWAP) data (Raupach et al., 2009) described by Jones et al. (2009). The rainfall data on the 0.05° × 0.05° grid was smoothed to a 0.5° × 0.5° grid for analysis since the extra resolution is not needed here.

The classification of synoptic systems used here follows Pook et al. (2006). They defined an analysis box centred west of the Mallee region (Figure 1) to classify synoptic systems. Their scheme classifies rain days according to three basic synoptic types: ‘cold-frontal systems of all types, cold-cored lows that have become cut off from the westerly airstream (cutoff lows), and a combined category designated ‘others’, which includes particular airstream types, waves in the easterlies, and open troughs aloft’ (Pook et al., 2006). The classification of synoptic systems utilizes the mean sea level pressure field, the 500 hPa height field, and the 1000–500 hPa thickness field and is cross referenced against both Bureau of Meteorology analyses and data from the NCEP/NCAR reanalysis (Kalnay et al., 1996) over the period 1956–2009. The manual classification of cutoff low systems is also cross referenced against the results from an automatic recognition system for cutoff lows as an additional quality control measure.

The synoptic analysis covers the period 1956–2009 when we have high confidence in the classification of synoptic systems in the southeastern Australia region. This is a longer period than that used to characterize rainfall deciles for the recent drought, which is 1996–2009. Use of the longer period allows us to place synoptic system behaviour during the drought in the context of variation of synoptic systems in the preceding four decades. In some cases, we have calculated trend rates for key variables. The trend plots are based on data calculated from 1978. That choice was made to ensure a reasonable sample of data before, as well as during, the drought to show the changes. If the trend were calculated over the period of the drought only, then we may not see the changes clearly as the change may have largely already occurred.

In the analyses that follow, we show rainfall trends for the April–October period (the period for which we have synoptic classifications), which encompasses much of the rainfall in the Mallee region. There is some seasonality to the rainfall decline in the southeast, which has been particularly pronounced in autumn (Murphy and Timbal, 2008). The major declines in southeast rainfall occur from March to October (Timbal, 2009), which matches well with the period of analysis. We also calculated results for the separate subperiods (April–June, June–August, August–October), which show minor variations from the full April–October results.

The analysis makes use of a blocking index and a measure of baroclinicity. The blocking index developed by the Australian Bureau of Meteorology for use in the Australian region is defined as follows: blocking = 0.5(U25 + U30U40 − 2U45U50 + U55 + U60) where Uy represents the zonal component of the 500 hPa wind at latitude y (Pook and Gibson, 1999). Daily wind data from the NCEP/NCAR reanalysis (Kalnay et al., 1996) is used to calculate the blocking index. The measure of baroclinicity used here is the Eady growth rate. This index assesses baroclinic instability through the vertical gradient in horizontal wind in the middle troposphere (Hoskins and Valdez, 1990). The Eady growth rate, σBI, is defined by: equation image where f is the Coriolis parameter, N is the Brunt-Väisällä frequency, z is vertical distance, and v is the horizontal wind vector at 400 and 600 hPa (Paciorek et al., 2002). This provides a measure of baroclinicity centred about 500 hPa, which is appropriate for assessing the baroclinicity of midlatitude systems.

3. Rainfall and synoptic trends

The drought of 1996–2009 covered parts of Queensland and southwest Western Australia, but has largest spatial extent in the southeast quadrant of Australia as shown in Figure 3 for April–October rainfall. Much of the southeast region, including the Mallee, experienced the lowest period of rainfall in the instrumental record. The instrumental record has consistent coverage from about 1900 onward.

Figure 3.

Rainfall deciles of April–October rainfall for the period 1996–2009. The decile rankings are based on all 14-year periods in the gridded data spanning the years 1900–2009. Note that the extremes of the colour bar refer not to deciles but to the lowest and highest 14-year averages in the record and are enclosed by the thicker contour line on the plot

The time series of Mallee rainfall for the April–October months is shown in Figure 4. The total rainfall curve (black) is fairly flat from the 1960s through the mid-1990s and then drops substantially, as for other locations in the southeast (Fawcett, 2004). The drop in rainfall is reflected in each of the synoptic types (cutoff, frontal, stream), but is most pronounced for cutoff rainfall (blue curve). The multidecadal variability in total rainfall is also best matched by that in cutoff systems, which contribute most to total rainfall.

Figure 4.

Time series of April–October rainfall for the Mallee. The colors denote: black (total rainfall), blue (cutoff rainfall), red (frontal rainfall), green (stream system rainfall). The solid lines follow annual values and the dashed curves are loess smoothed fits to the series

The question arises whether the trends in rainfall from each type of synoptic system are driven by changes in the frequency and/or intensity, or other characteristics of these systems. Figure 5a shows two measures of the frequency of cutoff systems through time in the southeast (the number of cutoff systems and the number of cutoff days each April–October). There is considerable multidecadal variability in the frequency of cutoff systems in the region, which matches well the multidecadal variability in the amount of rain from cutoffs (Figure 4, blue curve).

Figure 5.

Time series of (a) the frequency of cutoff low systems measured according to the number of cutoff systems (black/lower curve) and the number of days in which cutoff systems occur (blue/upper curve), and (b) the intensity of cutoff systems measured according to the amount of rain per cutoff system (black/upper curve) and the amount of rain per cutoff day (blue/lower curve). The data span the April–October period each year. The dashed curves are loess smoothed fits to the series.

There are a variety of possible ways to measure the intensity of cutoff systems. We chose the amount of rain per cutoff system (and per cutoff day) as it relates directly to rainfall variability, which is our main interest. These measures of intensity are shown in Figure 5(b). These indicators are less obviously related to multidecadal variability of rainfall than the measures of cutoff frequency in Figure 5(a). The smoothed fit line in Figure 5(b) shows a long period decline in the amount of rain per cutoff system (and per cutoff day).

Cutoff systems usually occur in association with blocking high pressure systems, which cause them to ‘cut off’ from the main westerly stream. Pook et al. (2006) and Risbey et al. (2009b) have shown that blocking at 140°E longitude is well correlated (+ve) with rainfall in southeastern Australia. We are interested in knowing further whether blocking influences the frequency or intensity of cutoff systems in this region, or both. The time series of blocking at 140°E is more highly correlated with the number of cutoff systems (r = 0.6) than the intensity of cutoff systems (r = 0.2), implying that the connection to cutoff rainfall is via the frequency of systems. Figure 6 shows series of the blocking index and frequency of cutoff systems. Multidecadal variability in the number of cutoff systems is reasonably well matched by multidecadal variability in blocking.

Figure 6.

Time series of the frequency of cutoff low systems measured according to the number of days in which cutoff systems occur (blue curve), and the blocking index at 140°E (red curve) [units m s−1]. The dashed curves are loess smoothed fits to the series

The power spectrum of blocking in the Tasman Sea region exhibits power at decadal to multidecadal periods, and also at shorter periods between 3 and 4 years (Figure 7). The 3- to 4-year peaks may reflect a response of blocking to the ENSO (Renwick, 1998), perhaps through the influence of Rossby waves propagating into the south Pacific region (Mo and Ghil, 1987). Spectral analyses of the series in Figures 4 and 5 (not shown) indicates that the power at multidecadal frequencies in total rainfall (Figure 4, black curve) is contributed mostly by power at multidecadal frequency in cutoff systems (Figure 4, blue curve). Given the strong association between blocking in the Tasman and cutoff systems in the southeast, it seems plausible that multidecadal variability in the number of cutoff systems would be related to the multidecadal variability in blocking. As such, blocking would be contributing to multidecadal variability in rainfall in the southeast, though a longer series of data would be needed to confirm results on these time scales.

Figure 7.

The plot shows the square root of the power of the blocking index. The power spectrum was calculated for each longitude between 100 and 300°E to produce a contour plot as a function of longitude and period. The power spectrum was calculated for the April–October months for the period 1956–2009.

Frontal systems exhibit less apparent variability at decadal periods than cutoff systems (Figure 8(a)). As for cutoff systems, there is a long-term downward trend in the amount of rain per frontal system (Figure 8(b)).

Figure 8.

Time series of (a) the frequency of frontal systems measured according to the number of frontal systems (black/lower curve) and the number of days in which frontal systems occur (blue/upper curve), and (b) the intensity of frontal systems measured according to the amount of rain per frontal system (black/upper curve) and the amount of rain per frontal day (blue/lower curve). The data span the April–October period each year. The dashed curves are loess smoothed fits to the series.

We are now in a position to decompose the change in rainfall in the Mallee region into contributions from cutoffs and fronts. The smooth fit to the rainfall series in Figure 4 shows a decline from just after 1990, continuing through the recent drought. There is a decline in the amount of rain from both cutoffs and fronts over this period, and this in turn is composed of reductions in the frequency and intensity of these systems. Table 1 shows how the total decline in rainfall since 1990 is partitioned into contributions from changes in the frequency and intensity of systems. About two-thirds of the recent decline is related to the reduction in cutoff rainfall, which in turn is about evenly due to reductions in the number of cutoff systems and in the rain per cutoff system. About a third of the recent decline is due to frontal systems, but this is mostly due to a reduction in the amount of rain per frontal system. The reduction in rainfall since 1990 is composed of reductions in both the frequency and intensity of systems, though the reduction in intensity (rain per system) accounts for the largest share.

Table 1. Contributions to the roughly 105mm decline in April–October Mallee total rainfall from 1990–2007. The contributions are from the decrease in number of cutoff systems (times the average amount of rain per cutoff system), the decline in rain per cutoff system (times the average number of cutoff systems), the decline in number of fronts (times the average amount of rain per frontal system), and the decline in rain per frontal system (times the average number of frontal systems)
Cutoff frequency− 5 cutoff systems × 7 mm/cutoff system = 35 mm
Cutoff intensity− 2 mm/cutoff system × 18 cutoff systems = 36 mm
Frontal frequency− 2 frontal systems × 2.5 mm/frontal system = 5 mm
Frontal intensity− 1 mm/frontal system × 30 frontal systems = 30 mm

The changes in synoptic components of rainfall described above have some implications for the kinds of circulation changes that might have caused them. First, the decline in number of frontal systems (and frontal days) is fairly weak and accounts for only a small part of the overall rainfall decline. The frequency of fronts in southeastern Australia is related to the location of the Southern Hemisphere storm tracks and subtropical ridge over the continent. The mean latitude of the subtropical ridge and storm tracks migrates with the seasons, allowing more frontal passages in the cool seasons when the subtropical ridge is closer to the equator. On decadal time scales, there is evidence for an intensification and spreading of the subtropical ridge (Timbal, 2009) and a poleward displacement of storm tracks (Frederiksen and Frederiksen, 2007). These trends could have reduced the frequency of frontal passages, though the front frequency contribution is a relatively minor one to the overall rainfall trend in the southeast.

On the other hand, the reduction in frequency of cutoff systems is a significant component of the overall rainfall decline. The frequency of cutoff systems in southeast Australia is related in part to blocking activity in the Tasman Sea region, which shows a mean decline over the period since 1990 (Figure 6). In order to understand this component of the rainfall decline, we need to understand the processes driving blocking variability in the region. Issues related to the change in frequency and intensity of systems are taken up in the following sections.

4. Changes in system intensity

The observed reduction in the amount of rain per cutoff and frontal system over the period since 1990 could have a variety of different causes. Less rain per system could be a consequence of:

  •  A reduction in the intensity of systems. Weaker synoptic systems would generally produce less rain (Risbey et al., 2009a). Frederiksen and Frederiksen, (2007) note a general reduction in the baroclinicity of systems in the Southern Hemisphere storm track, persisting through the present drought (Frederiksen et al., 2009). Such a reduction in baroclinicity of systems would result in less rain per system (Risbey et al., 2009a). A reduction in intensity of frontal systems could be associated with a poleward shift in latitude of cyclones and fronts such that the weaker equatorward portion of the frontal band passes through the region (Simmonds and Keay, 2000b).

  •  A change in development stage over southeastern Australia. Synoptic systems tend to produce more rainfall during the mature stage of the lifecycle than at onset. A change in the preferred locations of cyclogenesis and cyclolysis of systems (Simmonds and Keay, 2000a) could result in a change in typical lifecycle stage over southeastern Australia. To date such trends have not been examined with regard to the recent drought.

  •  A decrease in moisture entrainment. Synoptic systems could produce less rainfall if there is a change in the typical moisture pathways into the system. McIntosh et al. (2007) show that rainfall in cutoff systems is associated with a variety of different moisture pathways from the north and south of the continent. The highest rainfall events generally occur when moisture is entrained from the marine boundary layer off northeastern Australia. A change in the tendency for systems to entrain moisture from this region could produce a reduction in rain per system.

  •  Increase in speed of systems. Cutoff systems produce more rainfall than frontal systems (more than twice as much per system in the southeast), in part because they travel more slowly through the domain and provide more concentrated rainfall over a longer period. An increase in the speed of cutoff or frontal systems, would (ceteris paribus) produce a reduction in rainfall per system in the study region. Lim and Simmonds (2007) find some evidence for an increase in the speed of Southern Hemisphere extratropical cyclones over 1979–2001. A more targeted study focusing on the study area and period would be needed to quantify the contribution of cyclone speed to rainfall changes in southeastern Australia.

The resolution of which of these (or other) factors is driving the observed reduction in rain per system requires further work to quantify the influence of each factor, which is beyond the scope of this study. We explore further here only the first factor, which is the role of the possible reduction in intensity of systems.

Frederiksen and Frederiksen (2007) and Frederiksen et al. (2009) show a reduction in storm track instability modes growing on the subtropical jet in recent decades. They note that there has been a decrease in subtropical jet zonal winds throughout most of the Southern Hemisphere. This decrease is evident in seasonal mean fields in Australian longitudes, though winds on the polar side of the mean jet position have increased slightly. If the seasonal mean changes are representative of the jet changes accompanying storm systems, then we would expect a weakening of storm intensities.

As a measure of changes in system intensity, we show the trend in Eady growth rate [a measure of baroclinic instability; Hoskins and Valdez, 1990] for cutoff and frontal systems in Figures 9 and 10 respectively. The Eady trends are calculated over the period 1978–2007 to include the recent drought and a preceding period. The trends in Eady growth rate over the Australian region are generally small over the period examined. The key region from the perspective of the systems examined here is over the synoptic box where Eady growth rates are a maximum in the mean Eady growth rate fields in conjunction with the subtropical jet. For cutoffs there is a weak decreasing trend in Eady growth rate over all cutoff events (Figure 9(a)), and a mixed result with regions of positive and negative Eady trends when the smallest cutoff events are excluded from the composite (Figure 9(b)). As such, the Eady growth rate does not provide a clear guide to intensity changes in cutoff lows.

Figure 9.

Linear trend in Eady growth rate for cutoff systems in April–October 1978–2007. The trend is calculated over all days in which a cutoff system is diagnosed in the synoptic box indicated by the rectangular box (as in figure 1). The left panel is the average trend over all cutoff days. The right panel is an average over just the subset of cutoffs that produced rain days between 5 and 20 mm across the Mallee region. Dash contours indicate negative trends and solid contours indicate positive trends

Figure 10.

As in figure 9 but for Eady growth rate linear trend for frontal systems

For fronts the Eady growth rate trends are also weak for a composite over all fronts in the period (Figure 10(a)). When the weaker fronts are excluded and only those in the 5–20 mm d−1 category remain (Figure 10(b)), the Eady growth rates decline across the synoptic box and the midlatitude zone. The latter is more consistent with expectations from the storm track studies cited above. The center of the low in frontal systems is more coherently associated with the storm tracks than cutoff low systems, which tend to occur equatorward of the main storm tracks. Thus, it is not surprising that the intensity of fronts tends to weaken by the baroclinicity measure.

In order to assess the relative importance of trends for systems of different rainfall intensity, we show the contributions to total cutoff and frontal rainfall from small (<5 mm d−1), medium (5–15 mm d−1), and large (>15 mm d−1) events in Figure 11. This figure illustrates that cutoff rainfall is dominated by medium size events in the 5–15 mm category, while frontal rainfall is usually dominated by small events in the < 5 mm category. However, in the period since 1990 when rainfall has declined, the decline is mostly associated with rainfall in higher storm size categories. For cutoff rainfall it is the > 15 mm category that shows the biggest decline, and for fronts it is the 5–15 mm category that declines, not the dominant < 5 mm category.

Figure 11.

Time series of the amount of rain over April–October each year from (a) cutoff systems and (b) frontal systems. The black curve is the total amount of rain from these systems. The blue curve labelled ‘small’ is the amount of rain in the band 0–5 mm, the red curve labelled ‘medium’ is the amount of rain in the band 5–15 mm, and the green curve labelled ‘large’ is the amount of rain from events greater than 15 mm. The dashed curves are loess smoothed fits to the series

The decline in cutoff rainfall in the > 15 mm category is mostly a consequence of a drop in frequency of intense cutoffs. The number of the large rainfall cutoff systems shows a marked drop in the most recent decade marking the drought (Figure 12(a)). The average amount of rain per high rain cutoff day is higher in the first half of the record (1956–1982) than in the second half (1982–2009) in Figure 12(b), but there is no particular trend in the intensity of high rain cutoffs immediately preceding or during the drought.

Figure 12.

Time series of (a) the number of cutoff rain days of greater than > 15 mm/day rainfall each April–October (with loess smoothed fit dashed) and (b) the average amount of rain per day for cutoff systems of greater than > 15 mm/day rainfall each April–October. Zero values correspond to years in which there are no cutoffs in this intensity range.

The decline in frontal rainfall in the 5–15 mm category is mostly a consequence of the decline in the number of systems in this intensity range during the period of the drought (Figure 13(a)). There is no apparent trend in the intensity of frontal systems within this band (Figure 13(b)). A weakening of intensity of frontal systems in general during the drought (Table 1 and Figure 8) is consistent with a reduction in numbers of the medium and high intensity frontal systems in favour of weaker systems.

Figure 13.

Time series of (a) the number of frontal systems of between 5 mm/day and 15 mm/day rainfall each April–October (with loess smoothed fit dashed) and (b) the average amount of rain per day for frontal systems of between 5 mm/day and 15 mm/day rainfall each April–October. Zero values correspond to years in which there are no fronts in this intensity range.

5. Changes in system frequency

In Section '3. Rainfall and synoptic trends', we noted that about half the contribution of the reduction in cutoff rainfall since 1990 is due to the reduction in frequency of cutoff systems [and this mostly for larger cutoff systems (Section '4. Changes in system intensity')]. Furthermore, the reduction in frequency of cutoff systems is associated with a reduction in blocking in the Tasman Sea region. In this section, we examine the changes in blocking in broader hemispheric context to probe further the reasons for the drop in cutoff frequency. This section is concerned only with changes in cutoff frequency, not fronts, as the contribution of changes of frontal frequency to the recent drought rainfall decline is very small (Table 1).

Cutoff systems in southeast Australia usually form in conjunction with blocking highs in the Tasman Sea region. The Tasman/New Zealand longitude sector is the most preferred region for Southern Hemisphere blocking (Trenberth and Mo, 1985; Pook and Gibson, 1999). Blocks in this region form in association with split upper level flow (Bals-Elsholz et al., 2001) and a ridge in the longwave pattern (Coughlan, 1983), particularly in response to topographic forcing of the flow field (Frederiksen, 1984; Zidikheri et al., 2007). The tendency for split upper level flow in the region is enhanced during the cooler season by the anomalous easterly thermal wind forced by the cold Australian continent and the relatively warm ocean to its south (Taljaard, 1972; Pook, 1994).

While blocks may prefer the Tasman region, they do occur around the hemisphere and reinforce a three wave pattern in the Southern Hemisphere. Further, blocks exhibit long time scale variability in their precise locations about the preferred regions. In Figure 14, we show a Hovmoller plot of the blocking index over the eastern Indian Ocean, Australian, and Pacific Ocean region. The most active centre of blocking shifts longitudinally from year to year and decade to decade (as indicated by the thick dashed line in Figure 14 which tracks the longitude of the maximum value of the blocking index each year). The longitude of maximum blocking index varies over a wide range between about 150°E and 220°E over the period. Years when Pacific blocking activity shifts westward toward the Tasman Sea tend to be associated with higher rainfall in southeast Australia, whereas years when blocking shifts further eastward into the Pacific tend to be dryer there.

Figure 14.

Hovmoller plot of the blocking index [m s−1] for the April–October period as a function of time (1956–2009) and longitude (100–300°E). The thick dashed line marks the longitude at which the blocking index (averaged over the April–October period) is a maximum each year

There is considerable decadal variability apparent in blocking activity in Figure 14, with relatively more quiescent blocking periods in the beginning and end of the record shown. The decadal periodicity in blocking is evident in Figure 7, which shows spectral peaks at decadal frequencies and beyond.

Given the long period variability in blocking, one must use caution in fitting a linear trend over shorter periods. We are interested in the decline in blocking in the period of the recent drought (1996–2009) at the end of the record. We highlight this decline by fitting a linear trend to the blocking index at each longitude in a period preceding and including the recent drought. The trend fitted since 1978 is only intended to illustrate the recent decline in blocking and is not indicative of longer term trends. Figure 15 shows the result of fitting a linear trend to the blocking index at Australian and Pacific longitudes (100–300°E). In the period since 1978 there is a decline in the magnitude of the blocking index in the main Pacific blocking sector between 140°E and 240°E. The trends are positive only in the regions outside the main blocking sector. The trends are only weakly significant for the April–October period. The significance of the decline in blocking is higher if only the winter period is examined. From the perspective of southeast Australia, the relevant blocking region is the Tasman Sea section between 140°E and 160°E, which falls within the region of blocking index decline.

Figure 15.

Linear trend of the blocking index for the April–October period as a function of longitude (100–300°E). The circles indicate the mean best fit linear trend estimate at each longitude. The plus signs show the 10th percentile and 90th percentile confidence limits on the trend estimate from a bootstrap resampling of the blocking series. The trend fit is for the period 1978–2009 and is not representative of longer periods or different periods.

The decreasing trend in blocking activity in east Australia and the Tasman Sea region is consistent with the decrease in frequency of cutoff systems in southeast Australia. This provides a proximate explanation for the decrease in cutoff rainfall observed, but leaves open the larger questions of whether other processes contributed to this decline, and what controls the trends in blocking in the region. Like midlatitude eddies, blocking is generated by baroclinic instabilities (Frederiksen, 1984; Frederiksen and Frederiksen, 2011). It is possible that the decline in blocking is related to the observed reduction in baroclinicity poleward of the Australian continent (Frederiksen and Frederiksen, 2007; Frederiksen et al., 2009). Some evidence for this link is provided in Figure 9b. This figure shows the Eady growth rate (baroclinicity) trend for days on which all but the smallest cutoff systems occur. Since cutoff systems in the southeast generally occur in conjunction with blocking in the Tasman region, we can examine the trends in baroclinicity in the Tasman region as it would relate to blocking events. The baroclinicity uniformly declines across the Tasman region in Figure 9b. This is consistent with the observed reduction of blocking in this region and with the results of Frederiksen and Frederiksen (2011), who find a reduction in the growth rate of blocking modes in the recent drought. Frederiksen and Frederiksen (2011) in turn relate the reduction in blocking growth rates to a reduction in subtropical jet intensity and baroclinicity.

6. Conclusions

This study has examined the contributions of different synoptic types to rainfall variability and trends in southeast Australia. The main focus has been on the drought of 1996–2009, which is characterized by the lowest fourteen year rainfall totals in the century long record.

The reduction in rainfall during the recent drought is mainly composed of reductions in rainfall from cutoff low systems and frontal systems. Cutoffs account for about two-thirds of the reduction and frontal systems account for about one-third of the reduction. Cutoff systems are more closely associated with blocking than with other features, while frontal systems are more closely associated with the latitudinal positions of the subtropical ridge and storm tracks. The centres of the lows spawning fronts sit in the Southern Hemisphere storm track, and the fronts tend to span the region from the low centre equatorward to the poleward edges of the subtropical ridge.

Since cutoffs dominate fronts in accounting for the rainfall reduction, explanations for the recent drought must account for the processes controlling cutoff low systems. Thus, for example, explanations that centre around the subtropical ridge apply mainly to fronts and do not readily provide a complete explanation of the drought.

The reduction in rainfall from frontal systems during the drought is mostly due to a reduction in the amount of rain per frontal system, rather than to a reduction in the overall number of frontal systems. The reduction in rain per frontal system is associated with a reduction in the number of fronts in the 5–15 mm d−1 category. The reduction in numbers of more intense fronts is consistent with previous work on the reduction in baroclinicity in the storm tracks south of Australia (Frederiksen and Frederiksen, 2007). This work documents a reduction in mid-tropospheric Eady growth rate (baroclinicity) across southern Australia in association with the more intense frontal systems. The explanation that the observed reduction of baroclinicity and contraction of Southern Hemisphere storm track is causing a reduction in rainfall is consistent for the one-third frontal contribution to the recent drought.

The two-thirds reduction in rainfall from cutoff systems during the drought is due to a decline in the frequency and intensity of cutoff systems. Specifically, there is a reduction in the number of the most intense cutoff systems. The frequency of cutoff systems in the region seems to be related to variability in blocking in the Tasman region, which presents a mechanism to influence multidecadal rainfall variability in southeast Australia. Blocking has undergone a weak decline in the Tasman region over the drought period and thus provides a potential proximate explanation for the decline in cutoff rainfall during the drought. The decline in blocking itself is associated with broader reductions in baroclinic instability in midlatitudes. Analysis of baroclinicity in the Tasman blocking region on days when cutoff systems occur is consistent with this in showing a decline in baroclinicity in the Tasman then. As such, the broader decline in baroclinicity in the storm track could be associated with the two-third cutoff contribution to the drought as well.

The general picture of the recent drought that emerges here is of a drop in rainfall dominated by cutoff systems. That drop appears in the context of multidecadal variability of cutoff rainfall and blocking in the region. A reduction of blocking in the Tasman Sea region seems to be driving a reduction in the number of intense cutoff systems during the drought. The reduction in blocking is coincident with a reduction in baroclinicity in the Tasman region, which in turn may be related to broader scale reductions of baroclinicity in this latitude zone.

This study has highlighted the manner in which synoptic classifications can shed light on the proximate causes of rainfall variations and trends. By knowing which synoptic systems are associated with which circulation features, it is possible to start ruling out some postulated causes of observed rainfall trends. This method can only provide a partial explanation of trends, but that may still be an advance on existing understanding.

Acknowledgements

This work was funded by the Climate Adaptation Flagship of CSIRO and the Managing Climate Variability program of the Grains Research and Development Corporation, Australia. We are grateful for the constructive reviews.

Ancillary