A pressure gradient metric capturing planetary-scale influences on eastern Australian rainfall



[1] The Gayndah-Deniliquin index (GDI), a measure of the north-south atmospheric pressure gradient across eastern Australia, is presented. The 113 year long GDI record reveals strong interannual to decadal scale variability in zonal geostrophic wind flow across eastern Australia. The GDI, as a measure of easterly geostrophic wind strength and associated moisture transport from the Pacific Ocean, is shown to be significantly correlated with summer rainfall over vast areas of the Australian continent, especially over the Murray Darling Basin and the state of New South Wales. The latest abrupt decline in the GDI, which commenced around 2001, corresponded with the onset of a severe prolonged drought across eastern Australia. We demonstrate that the northern and southern poles of the MSLP derived GDI are differentially influenced by El Niño-Southern Oscillation (ENSO) and the Southern Annular Mode (SAM). Understanding the effects of these interactions between SAM and ENSO on moisture transport to eastern Australia could have implications for future Australian climate variability and climate change.

1. Introduction

[2] Considerable attention is now being given to the possible impacts of anthropogenic climate change on eastern Australian rainfall following the observed abrupt decline in annual rainfall across much of central-eastern Australia since the year 2000, and further south (in the state of Victoria) since 1996 [Murphy and Timbal, 2008]. The current protracted drought across eastern Australia has severely impacted agricultural productivity, depleted water storages to critical levels, and resulted in serious reductions in environmental river flows. Five year inflows to the important Murray Darling Basin (MDB) (Figure 1), which accounts for the vast proportion of irrigated agricultural production in Australia, have been the lowest on record (since 1891) during the period 2001 to 2006 [Matthews, 2006]. The recent abrupt rainfall decline is contextualized by historical river observations which show abrupt shifts repeatedly occurring between multidecadal drought and flood dominated periods along the east Australian coast [Erskine and Warner, 1988]. However, multidecadal drought dominated periods are poorly understood with causes of the shifts between high rainfall and drought phases remaining obscure [Vives and Jones, 2005].

Figure 1.

Schematic of the study region showing the location of the two main MSLP observation sites (Deniliquin and Gayndah). The mean positions of the subtropical ridge (STR) of high pressure (in summer and winter) and the semi-permanent summer eastern and western low pressure troughs are also shown.

[3] Meinke et al. [2005] have shown that decadal scale variability of annual rainfall affects much of central-eastern Australia while multidecadal rainfall variability mainly affects southeastern Australia. Previous studies of rainfall variability in eastern Australia have explored relationships with El Niño Southern Oscillation (ENSO) [Power et al., 1999; Verdon et al., 2004], Indian Ocean Dipole (IOD) [Power et al., 1999; Meyers et al., 2007], Southern Annular Mode (SAM) [Gillett et al., 2006; Hendon et al., 2007], mean sea level pressure (MSLP) and various indices used to measure the latitudinal position of the subtropical ridge (STR) of high pressure [Drosdowsky, 2005]. There remains, however, a relatively poor understanding of the synoptic scale atmospheric mechanisms that contribute to the observed east Australian warm season multidecadal rainfall variability.

[4] This study investigates the relationship between the austral summer easterly geostrophic wind flow velocities across the state of New South Wales (NSW) and interannual to multidecadal rainfall variability across Australia. The variability in easterly geostrophic flow is captured using a new index representing the meridional atmospheric pressure gradient changes across NSW.

2. Data and Method

[5] We use the Australian Bureau of Meteorology (BoM) 0.25° by 0.25° gridded rainfall dataset generated using the Barnes interpolation method [Jones and Weymouth, 1997], as well as annual and monthly NSW state-wide averaged rainfall data provided by the BoM.

[6] To explore relationships between Australian rainfall variability and changes in the synoptic scale easterly flow we use the meridional atmospheric pressure gradient across NSW as a simple measure of the geostrophic wind velocity. This avoids the limitations of using surface wind observations on smaller scales when local wind effects (shallow sea breezes, anabatic/katabatic effects and topographic influences) become more important and are not considered relevant to this study.

[7] Monthly-averaged 9 am local time (+10 GMT) MSLP data for two locations were used in this study: Gayndah (25.6°S, 151.6°E) in southeast Queensland (QLD) and Deniliquin (35.6°S, 144.9°E) in southern NSW. These sites, separated by some 1270 km, were chosen because they meridionally span the MDB and NSW (Figure 1), importantly they form a transect which is roughly parallel with the NSW coast and the Great Dividing Range, and because they provide long contiguous records with very few missing months of summer data, with replacement MSLP data sourced from surrounding stations.

[8] For the purposes of this study, a new index, the Gayndah – Deniliquin Index (GDI) is used to represent the meridional sea level atmospheric pressure gradient - a proxy for the large scale zonal geostrophic wind flow (Vg, ms−1) across temperate eastern Australia between 25°–35°S (Figure 1):

equation image

where g is gravity (ms−2), f is the Coriolis parameter (s−1) and δz/δn is the slope of the isobaric surface normal to the contour lines to the right of the direction of motion (southern hemisphere). Put simply:

equation image

Positive (negative) values of the index are indicative of predominantly easterly to south-easterly (westerly to north-westerly) geostrophic wind flow across the state of New South Wales. The GDI was calculated for every month from September 1893 until December 2006. The index has a marked annual cycle with a maximum in March (+2.55 hPa, averaged over 113 years) and a minimum in September (−1.25 hPa, averaged over the same period). Seasonal variations in the GDI correspond with seasonal rainfall variations at many coastal locations in northern NSW; climatological rainfall maxima and minima are in March and September respectively, matching the GDI.

[9] Monthly values of the Troup [1965] Southern Oscillation Index (SOI) and Darwin MSLP (from the BoM; http://www.bom.gov.au/climate/current/soihtm1.shtml), the Multivariate ENSO Index (MEI), Trans-Niño Index (TNI), Niño3.4 and Niño4 sea surface temperatures (SST) (from NOAA; http://www.cdc.noaa.gov/ClimateIndices/) were used. Monthly values of the Marshall [2003] SAM index and the Meyers et al. [2007] IOD index were also used in this study. Correlations between climate indices and rainfall in this study take account of serial correlation in the time series in terms of effective degrees of freedom according to Davis [1976]. All time series in this study were linearly detrended before performing the correlation analyses.

3. Results

[10] Interannual to multidecadal variations in the December-January-February (DJF) GDI show a strong relationship with observed DJF rainfall variability across much of eastern Australia, especially over NSW (Figures 2a and 2b). The correlation coefficient between NSW state-wide summer (DJF average) rainfall and the summer GDI is r = 0.47, significant at the 99% level (sig.) (Figure 2a). Five-year moving averages were used to smooth the interannual peaks in the series to aid interpretation of the background decadal variability. The correlation coefficient between the smoothed time series of NSW state-wide summer rainfall and the summer GDI is r = 0.70, 99% sig. (Figure 2b). The smoothed time series of NSW state-wide annual rainfall is also strongly and significantly correlated (99% sig.) with the summer GDI, r = 0.67 (not shown). The highest GDI values are recorded in the 1950s and 1970s which correspond to the wettest decades in NSW during the 20th century. The summer GDI is shown to increase abruptly in the mid 1940s and again in the late 1960s to early 1970s, indicating marked increases in the easterly geostrophic wind velocities across 25°S–35°S at this time. Notably, the summer GDI and eastern Australian rainfall are both seen to shift abruptly and almost concurrently within the record.

Figure 2.

(a) Summer (DJF) GDI (dashed line) and DJF NSW state-wide average rainfall (solid line) (r = 0.47, 99% sig.) (b) Five-year moving averages of both the DJF GDI (dashed line) and DJF NSW state-wide average rainfall (solid line) (r = 0.70, 99% sig.).

[11] The highest value of the five-year moving average DJF GDI occurred in 1974 which is also the wettest year ever recorded for the Australian continent (based on BoM records which commenced in 1900). The GDI also highlights the dramatic decline in strength of the summer easterly geostrophic wind in recent years. The most recent smoothed GDI values are significantly lower than the lowest values recorded in 1947 at the end of the last multidecadal dry phase (Figure 2b). Interestingly, no more than two consecutive negative GDI years had been observed (1896/1897 and 1923/24) until the most recent four consecutive years of negative values, 2004 to 2007 (Figure 2a). The sharp rainfall decline in central-eastern Australia commencing in 2000/2001 coincides with the recent rapidly falling summer GDI (Figure 2b).

[12] To assess the strength of the unsmoothed relationship between the summer GDI and the summer season rainfall across Australia, zero lag correlation coefficients were calculated between the summer GDI and summer rainfall across all Australian grid cells. A broad swathe of statistically significant (>95% confidence level) correlations were found across eastern Australia (Figure 3a). The correlation is strongest (r > 0.5) along the heavily populated NSW – southern QLD coastline between Sydney and Brisbane. Statistically significant correlations >0.2 are evident across the MDB.

Figure 3.

Spatial correlations for DJF rainfall and (clockwise from top left) (a) DJF GDI, (b) DJF SOI, (c) DJF SAM and (d) DJF NINO4 indices. The catchment of the Murray Darling basin (MDB) is shown. All colored grid points display a relationship with >95% significance.

[13] To further investigate the utility and complementarity of the GDI, we compare similar synchronous correlation coefficient spatial distributions between DJF rainfall and commonly used ENSO indices (Figures 3b and 3c). The GDI is found to have a largely significant and more spatially extensive relationship with summer rainfall across the MDB and the east coast of Australia than all of the tested ENSO indices (Niño4, Niño3.4, MEI, SOI and TNI) and the IOD index. The monthly averaged DJF values of the Marshall [2003] SAM index were also used and display significant positive correlations with summer rainfall over much of the east and west of the continent (Figure 3d). Employing daily values of the SAM index, Hendon et al. [2007] report a similar influence of the positive phase of the SAM being related to increasing Australian rainfall, particularly summer rainfall in southeastern Australia.

[14] The relationship between the summer GDI and ENSO indices was investigated. There is a weak but significant relationship between the summer GDI and DJF SOI (r = 0.38, 99% sig.). Moderate correlations exist between the summer GDI and other tested Niño indices, the strongest of which was the DJF Niño 4 index (r = −0.55, 99% sig.).

[15] During the period 1899 to 1946 individual MSLP variations at Deniliquin and Gayndah were roughly synchronous (Figure 4b). In contrast, during the late 1940s and early 1950s, a strong rise in Deniliquin MSLP combined with a fall in MSLP at Gayndah resulted in a rapid rise in the GDI. The resulting rapid amplification of the meridional pressure gradient across NSW corresponds to the abrupt 1946/47 climate shift from dry to wet regimes in this region. The five-year running mean of Gayndah MSLP has been increasing since the mid-1970s concurrent with an increasing prevalence of El Niño events and decreasing SOI [Power and Smith, 2007]. The recent sharp change towards negative GDI values is the result of the continued rise of summer MSLP at Gayndah combined with a sudden decline of summer MSLP at Deniliquin since 2000.

Figure 4.

(a) Spatial correlation between summer (DJF) GDI and summer rainfall, following smoothing with a 5 year moving average. All colored grid points display a relationship with >95% significance. (b) Five-year moving averages of summer (DJF) MSLP at Gayndah, QLD (dashed line) and Deniliquin, NSW (solid line). (c) Five-year moving averages of both the DJF Deniliquin MSLP (solid line) and DJF SAM index (dashed line) (r = 0.70, 99% sig.). (d) Spatial correlation between DJF GDI and 850 mb relative humidity, 1949–2006 (NCEP).

[16] The decline in Deniliquin MSLP appears to be influenced by observed changes in the SAM. The five-year running mean of both the DJF SAM and Deniliquin MSLP peaked in 2000/2001 and have been in strong decline since (Figure 4c). The unsmoothed Deniliquin DJF MSLPs are correlated with the SAM at r = 0.35, 98% sig. (five year smoothed time series, r = 0.7, 99% sig., Figure 4c). The SAM has no significant relationship with Gayndah DJF MSLP (r = 0.02, <80% sig.). The SOI is strongly correlated with Gayndah MSLP (r = −0.57, 99% sig.), with a weaker but still highly significant relationship with Deniliquin DJF MSLP (r = −0.36, 99% sig.). Similarly, Niño4 DJF SSTs are strongly associated with Gayndah MSLP (r = 0.5, 99% sig.) but have no significant relationship with Deniliquin MSLP (r = 0.13, <80% sig.) Niño3.4 SSTs are also significantly related to DJF MSLP at the northern pole of the index, Gayndah (r = 0.57, 99% sig.), with a much weaker relationship at Deniliquin (r = 0.27, 90% sig.).

[17] NCEP reanalysis data display a strong relationship between the DJF GDI and relative humidity throughout the lower troposphere (surface-700 mb) over much of eastern Australia, particularly the MDB (Figure 4d). This is consistent with Stohl and James' [2005] analysis indicating that the Pacific Ocean is the dominant moisture source for rainfall over the MDB.

4. Discussion and Conclusions

[18] NSW is situated between 28°S and 37.5°S, just north of the mean climatological position of the southern hemisphere anticyclones. The mean seasonal position of the Subtropical Ridge shifts from 31°S (northern NSW) in late winter to 38°S (southern VIC) in late summer [Drosdowsky, 2005] (see Figure 1). The meridional pressure gradient north of the STR determines the strength of the easterly geostrophic wind flow across southern Australia in summer.

[19] The Gayndah-Deniliquin index (GDI) represents a measure of the strength of the zonal transport of either moist tropical maritime air (east to west) or dry subtropical continental air (west to east) across NSW/QLD latitudes. A high positive value of the GDI (>1 hPa) is indicative of sustained strong easterly zonal transport of Pacific maritime air onshore with resulting widespread high rainfall over eastern Australia. Conversely, a low value of the index (<1 hPa) is indicative of weak easterly winds and a greater predominance of westerly wind flow transporting dry continental air east across eastern Australia and offshore. The GDI is a clear indicator of the direction and relative strength of the zonal transport of these two adjacent air masses with highly disparate precipitable water contents.

[20] The trade wind regime over Australia is characterized by two marked troughs located on the western and eastern sides of the continent during summer (Figure 1). We contend that strengthening of the easterly geostrophic wind flow (higher GDI) increases moisture supply producing increased instability and rainfall associated with the eastern trough during these periods (Figure 4a). On decadal time-scales, periods of strong zonal easterlies (high values of the GDI) also appear to be related to drier conditions throughout the interior and southeast of Western Australia (Figure 4a).

[21] Gillet et al. [2006] and Hendon et al. [2007] have shown that the positive phase of the SAM increases MSLP over southern Australia, including over Deniliquin, the southern pole of the GDI (Figure 4c). The influence of the SAM on MSLP at the northern pole of the GDI, Gayndah, is largely reduced, where it is more influenced by ENSO in all seasons [Jones and Trewin, 2000; Trenberth and Caron, 2000]. Further, ENSO appears to make a similar contribution to the SAM on MSLP over the southern pole of the index. In summary, the poles of the GDI are forced to a greater extent by ENSO in the north and the combined influence of SAM and ENSO in the south during the austral summer. The GDI appears to combine the influence of these two planetary scale climatic drivers.

[22] The GDI, over interannual to decadal time-scales, has revealed abrupt changes in easterly geostrophic wind flow across NSW. A simple but effective diagnostic tool, the summer GDI captures the combined MSLP contributions from the SAM and ENSO to these abrupt changes in the easterly wind flow (and hence onshore/offshore moisture transports) across NSW. This study has shown that marked changes in geostrophic wind velocities are strongly related to rainfall changes in the Murray Darling Basin and over much of eastern Australia. The relationships between the SAM, ENSO, zonal winds and rainfall identified in this study may have implications for future eastern Australian climate change.


[23] C.S.R. is a Bureau of Meteorology staff member studying at Macquarie University, Sydney with the assistance of a BoM scholarship. B.T.'s contribution to this research is supported by the South Eastern Australian Climate Initiative (SEACI). Many thanks to Shayne McGregor (Macquarie University) for assistance with data analysis.