1. Introduction
The state of Queensland in northeastern Australia lies between the latitudes of 10° and 29°S and, with an area of 1.73 million km2, covers 22.5% of the land area of the Australian continent (Fig. 1). The World Heritage–listed Great Barrier Reef Marine Park extends for almost 2000 km along Queensland’s eastern coastline, from 10°40′ to 24°40′S. Although Queensland is not the largest Australian state, it has the largest area of rural holdings (1.58 million km2), the vast majority of which (1.51 million km2) comprises native and naturalized pastures grazed by cattle and sheep (Lloyd and Burrows 1988). The balance of rural holdings in Queensland comprises sown pastures and farming land, the former supporting mainly dairy cattle and the latter mainly dryland and irrigated cropping. Coal reserves occur throughout much of eastern Queensland with both underground and open-cut coal-mining operations undertaken, the latter accounting for approximately 85% of Queensland’s coal production. Beef and coal are Queensland’s main exports, accounting for almost half of Australia’s beef production (Department of Industry, Innovation and Science 2016) and just over half of Australia’s coal production (Meat and Livestock Australia 2016).
Queensland subregions: Queensland grazing lands (QGL) region, far-northern Queensland (FNQ), and far-western Queensland (FWQ). Locations mentioned in the text are indicated.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Queensland subregions: Queensland grazing lands (QGL) region, far-northern Queensland (FNQ), and far-western Queensland (FWQ). Locations mentioned in the text are indicated.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Queensland subregions: Queensland grazing lands (QGL) region, far-northern Queensland (FNQ), and far-western Queensland (FWQ). Locations mentioned in the text are indicated.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Depending on location, between 60% and 95% of Queensland’s mean annual rainfall occurs in the summer half of the year, from November to April (e.g., Dick 1958). Year-to-year variability in summer rainfall ranges from moderate to high by world standards (Dick 1958; Fatichi et al. 2012). The historical record shows long (5–10 yr) sequences of average to below-average rainfall, often broken by a cluster (2–5 yr) of summers with above average or extremely high rainfall (Day et al. 2005). Nearshore reefs carry the record of occurrence and intensity of freshwater flooding over many centuries, and this paleorecord is a reasonable proxy for statewide summer rainfall (Lough 1991). Like the contemporary rainfall record, this paleorecord is also characterized by strong year-to-year and multiyear variability, but it reveals even longer sequences of dry years (Lough 1991).
Interannual and multiyear fluctuations in summer rainfall show strong spatial coherence across Queensland, with objectively derived averages of statewide rainfall explaining approximately half the total variance in summer rainfall at any given recording station (Lough 1991). Since settlement of Queensland in the 1840s, severe drought has been experienced over a major part of the state in approximately one-third of years (Weston 1988). The first 17 years of the twenty-first century have seen a particularly high occurrence of drought in Queensland, with drought declarations extending over more than half of the state for almost 60% of the time. Widespread major flooding has occurred at irregular intervals throughout the climate record, most recently during the strong La Niña event in 2010/11. The series of events that constituted the “2010/11 Queensland floods” affected a large proportion of the Queensland population (Clemens et al. 2013), and resulted in almost all of the state being declared a natural disaster zone by the end of that summer. Although widespread rainfall was welcomed by the grazing industry, flooding and tropical cyclone activity caused major disruptions to the mining, construction, and tourism industries; widespread loss of crops; and widespread damage to infrastructure, roads, and property (Price Waterhouse Coopers 2011).
The strong spatial coherence of rainfall fluctuations in Queensland is due, in part, to the synoptic scale of rain-bearing systems. Easterly onshore flow is a major source of moisture for eastern Queensland and, with no substantial mountain ranges to interrupt the flow pattern, moist air tends to penetrate well inland (Nicholls 1992). During summer, the intertropical convergence zone (ITCZ) generally lies across far-northern Queensland at a mean latitude of approximately 18°S but, on occasions, can migrate as far south as 30°S, bringing widespread rainfall across Queensland (Sturman and Tapper 2006). Tropical cyclones develop over warm water to the north and east of Queensland, along both the ITCZ and the adjoining South Pacific convergence zone (SPCZ), and can also bring widespread rainfall, particularly if they make landfall and degenerate into slow-moving remnant low pressure systems. Planetary-scale phenomena, including the intraseasonal (approximately 40–50 day) Madden–Julian oscillation (Suppiah 1992) and the interannual El Niño–Southern Oscillation (ENSO; e.g., Allan 1988; Lough 1991; Suppiah 1992; Folland et al. 2002), contribute to the spatial coherence and variability of rainfall at monthly, seasonal, interannual, quasi-decadal, and multidecadal time scales, by influencing the position and strength of these synoptic-scale features and, in turn, the frequency and intensity of rain-bearing systems such as tropical cyclones (Callaghan and Power 2010).
Given the strong spatial coherence of rainfall fluctuations across Queensland, construction of rainfall time series for large portions of the state has been a useful first step in characterizing rainfall variability and analyzing the impact of ENSO and other climate drivers on Queensland rainfall (e.g., Walker and Bliss 1930; Treloar 1934; Rimmer and Hossack 1939; Lough 1991; Lough 1997; McKeon et al. 1998; Crimp and Day 2003; White et al. 2003; Risbey et al. 2009; Klingaman et al. 2013). Averaging rainfall on a seasonal basis, over a broad area, reduces the noise associated with both intraseasonal and local-scale rainfall variability. The resultant time series is therefore likely to better represent the more synoptic-scale, persistent, and potentially more predictable components of rainfall variability, a better understanding of which can be applied to improve the management of rainfall-related risks.
While several studies have produced statewide rainfall indices (e.g., Lough 1991, 1997; Klingaman et al. 2013), we have seen a need for a more regionally specific rainfall index. Almost half of the state, across Queensland’s far north and far west, is sparsely settled, carries relatively few livestock, and has low rainfall station density. Furthermore, far-northern and far-western Queensland represent opposite extremes of the state’s summer rainfall regime, the far north having high average summer rainfall totals with low year-to-year variability and, in contrast, the far west having low average summer rainfall totals with high year-to-year variability. Studies also indicate that the summer rainfall response to ENSO across far-northern and far-western Queensland differs from the remainder of the state (e.g., McBride and Nicholls 1983; Ropelewski and Halpert 1987; Lough 1991). Excluding these regions would therefore produce a more relevant index for the majority of the grazing industry in Queensland and, in turn, better reflect the historical impact of ENSO and other climate drivers at an industry level.
Hence, in the mid-1990s, we developed various rainfall time series based on a subregion of Queensland referred to here as the Queensland grazing lands (QGL) region (Fig. 1). Depending on the specific application, time series have been calculated either on an annual (April–March) basis, or for the summer pasture growing season, which we define here as extending from November to March. Rainfall time series based on our QGL region have proven their value within the following contexts: 1) documenting historical variability at the year-to-year, quasi-decadal, and multidecadal time scales that are most relevant to grazing or other applications (e.g., McKeon et al. 1998), 2) reporting the impacts of rainfall variability on animal production and the resource conditions of the grazing lands (e.g., McKeon et al. 2004; Stone et al. 2007), 3) understanding the historically varying link between rainfall variability and climatic drivers such as ENSO (e.g., McKeon et al. 1998; Crimp and Day 2003; White et al. 2003), 4) developing a climate risk assessment system to improve grazing management (Day et al. 2000), and 5) providing general advice to government and the community with regard to the impact of climate variability (e.g., Day et al. 2000; Stone et al. 2003). In addition to such applications, we consider the synoptic scale of these indices to be particularly suited to monitoring the emerging impacts of climate change (McKeon et al. 1998; Stone et al. 2007) and comparing the performance of various climate risk assessment systems, including GCMs (Syktus et al. 2003).
We consider our summer [November–March (NDJFM)] rainfall time series (above) to be the most appropriate single index for representing the rainfall variability faced by the majority of the grazing industry in Queensland. We refer to this index as the Queensland grazing lands rainfall index (QGLRI), which as defined here is a long-term (1890/91 summer onward) time series of summer (NDJFM) rainfall, spatially averaged over our QGL region (Fig. 1). Having developed our QGLRI in the mid-1990s, our primary aim here is to explain why, for the majority of the Queensland grazing industry at least, we consider our QGLRI to be a more appropriate index than statewide rainfall. Furthermore, by presenting our arguments here, we aim to increase awareness and acceptance of this index, and to also encourage and support any future studies of Queensland rainfall based on this index.
Our study is arranged in the following way. Section 2 describes the method used to construct our QGLRI, the data upon which the index is based, as well as data used elsewhere in our study. Sections 3 and 4 describe, from a grazing industry perspective, our choice to base our QGLRI on our QGL region (section 3) and our 5-month (NDJFM) summer season (section 4). In section 5, we show how our chosen region and season relate to airmass dominance over eastern Australia. In section 6, we compare the variability of our QGLRI with that of an equivalent statewide rainfall time series and assess how well these indices represent the summer rainfall variability across our QGL region. In section 7, we consider the location of our QGL region in relation to the spatially and temporally varying ENSO-rainfall response across Queensland. We summarize our study and present concluding remarks in section 8.
2. Data and methods
a. Queensland climate data
In this study we refer to region-wide averages of rainfall (sections 4–7), temperature (sections 4 and 5), and vapor pressure deficit (section 5). These calculations are ultimately based on raw data (monthly rainfall, daily minimum and maximum temperature, and vapor pressure) from the Australian Bureau of Meteorology’s (BoM) national observation network. These data are operationally (daily or monthly) sourced by the Queensland government and subjected to various automatic quality control checks (DSITI 2014a,b,c) prior to their incorporation into the Queensland government’s Strategic Information for Land Owners (SILO) climate database (https://silo.longpaddock.qld.gov.au/). In turn, these data are spatially interpolated across Australia (DSITI 2014a,b,c) and interpolated values are rendered onto a 0.05° × 0.05° spatial grid spanning the Australian region (112°–154°E, 10°–44°S). The number of contributing rainfall-recording stations across Australia has changed over time (DSITI 2014a) and, in Queensland, has risen from approximately 300 in the early 1890s to nearly 2000 in the early 1970s, with the current number standing at a little over 1000 stations (Fig. 2). By far the highest percentage (80%–85%) of Queensland stations are located in our QGL region, and this percentage has remained relatively constant over time (Fig. 2).
Number of Queensland rainfall recording stations reporting to the Australian BoM from 1870 to 2015. Numbers are broken down into the three Queensland subregions shown in Fig. 1, and a station is counted if reporting in November of a given year. Numbers prior to 1870 are not shown. Station openings and closures are available online (http://www.bom.gov.au/climate/data/).
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Number of Queensland rainfall recording stations reporting to the Australian BoM from 1870 to 2015. Numbers are broken down into the three Queensland subregions shown in Fig. 1, and a station is counted if reporting in November of a given year. Numbers prior to 1870 are not shown. Station openings and closures are available online (http://www.bom.gov.au/climate/data/).
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Number of Queensland rainfall recording stations reporting to the Australian BoM from 1870 to 2015. Numbers are broken down into the three Queensland subregions shown in Fig. 1, and a station is counted if reporting in November of a given year. Numbers prior to 1870 are not shown. Station openings and closures are available online (http://www.bom.gov.au/climate/data/).
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Derived variables (mean daily temperature and vapor pressure deficit) are also prepared on the same spatial grid as rainfall (above). Minimum temperature values are assumed to have occurred on the day of measurement, and daily maximum temperature values on the previous day. These data are realigned as such and, in accordance with common practice, the “average” daily temperature is calculated as the mean of the realigned values. This provides an approximation of the diurnal mean temperature that is consistent with the studies referred to in section 4 (McCown 1981a) and section 5 (Oliver 1970). The gridded data include the years 1889 to the present, and here we analyze data from 1890 to 2016.
Spatially averaged rainfall time series presented in this study were constructed by averaging values in each 0.05° × 0.05° grid cell across the three Queensland subregions shown in Fig. 1, as well as across the combined “statewide” region. We use the same procedure to derive region-wide averages of the temperature and vapor pressure deficit. A correction for decreasing grid-cell area with increasing latitude has not been applied to each 0.05° × 0.05° grid cell, and this is taken into account when calculating the average relative contribution of rainfall in each subregion to the statewide total (section 6). Our QGLRI, which we consider in detail here, is a time series of summer (NDJFM) rainfall totals, spatially averaged over the QGL region (Fig. 1). The index is available from the 1890/91 summer onward and is updated annually.
b. Other climate data
In section 7, we consider the correlation between the above rainfall time series and the Southern Oscillation index (SOI), values of which were provided by BoM (http://www.bom.gov.au/; accessed 17 February 2017).
c. Cadastral data
In demarcating the spatial boundaries of our QGL region (section 3), we reference the Queensland government’s Digital Cadastre Database (DCDB), which is the spatial representation of the property boundaries in Queensland and their related property descriptions. The DCDB is updated weekly and available from the Queensland Spatial Catalogue (Qspatial; http://qldspatial.information.qld.gov.au). Here, we present data downloaded in 2004 (Fig. 3a), but our original (mid-1990s) decisions, as outlined in section 3, were based on an earlier wall map derived from the same database (Skerman 1970).
Region of “previous” close land settlement in Queensland in relation to northern and western boundaries of our QGL region (straight lines: refer to Fig. 1): (a) boundary of region (solid line) superimposed on statewide cadastral mapping as in 2004 (the apparent shading is due to closeness of vectors delineating individual cadastres) and (b) the region of previous close settlement per se (shaded).
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Region of “previous” close land settlement in Queensland in relation to northern and western boundaries of our QGL region (straight lines: refer to Fig. 1): (a) boundary of region (solid line) superimposed on statewide cadastral mapping as in 2004 (the apparent shading is due to closeness of vectors delineating individual cadastres) and (b) the region of previous close settlement per se (shaded).
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Region of “previous” close land settlement in Queensland in relation to northern and western boundaries of our QGL region (straight lines: refer to Fig. 1): (a) boundary of region (solid line) superimposed on statewide cadastral mapping as in 2004 (the apparent shading is due to closeness of vectors delineating individual cadastres) and (b) the region of previous close settlement per se (shaded).
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
d. Livestock data
Because of difficulties in obtaining reliable livestock records at an appropriate scale, we have used property boundary information (above) to demarcate the spatial boundaries of our QGL region. However, to confirm our assumptions regarding a link between livestock distribution and property size, in section 3 we refer to an analysis of 2000/01 livestock data. The analysis, which translates Australian Bureau of Statistics (ABS) livestock statistics onto the same 0.05° × 0.05° resolution grid as climate data (above), is operationally conducted by the Queensland government to update the Australian Grassland and Rangeland Assessment by Spatial Simulation (AussieGRASS; Carter et al. 2003) spatial modeling system. The procedure is described by Carter et al. (1996), and involves heuristics to align differences in reporting criteria between each ABS census, to align different livestock classes on a “metabolic liveweight” basis, and to “redistribute” livestock units from statistical local areas (SLAs) to individual grid cells. Redistribution of livestock units is based on factors assumed to affect livestock distribution on a sub-SLA scale, including calculated pasture productivity, tree density, topography, and land-use restrictions on grazing.
3. Selecting geographic boundaries for our QGLRI
a. State boundaries
The current territorial boundaries of Queensland (Fig. 1) were demarcated in the mid-1800s when Queensland was still a colony of the British Empire. Administrators, at the time, had little knowledge of the future economic and production potential of agriculture in Queensland or of the potential limitations to agricultural expansion imposed by climate and rainfall variability. The southern boundary of Queensland was established in 1859, upon separation of the state of Queensland from the colony of New South Wales, and the current western boundary was established in 1862 when it was moved 3° of longitude to the west. Pastoralism and cropping had become well established in southern Australia by this time, but were only just becoming established in Queensland. Europeans were more accustomed, and their livestock and crops better adapted, to the temperate climate of present-day New South Wales than the tropical and subtropical environment of present-day Queensland. So it is by no coincidence that the current Queensland–New South Wales border aligns with the northern extent of the temperate zone. However, the northern and western limits of pastoralism in Queensland had not been “tested” at the time that the state boundaries were established, with only 25% of Queensland’s land area subject to grazing at that time (Ginn 2010). It would therefore be fortuitous if Queensland’s western boundary, or the state’s northern coastline, represent useful boundaries for delimiting climate zones related to livestock production.
b. Close land settlement
It took another 10 years (i.e., until 1870) for grazing enterprises to expand to all but the far north and far west of Queensland (Ginn 2010). The measurement of rainfall and other climate variables across Queensland then followed this expansion of land settlement. However, grazing potential was not truly tested until the 1920/30s when, in a bid to both increase agricultural production and to resettle returning soldiers and sailors after World War I (and again after World War II), Crown land was subdivided into “soldier settlement blocks,” each intended to sustain a family through agricultural activities. In most cases, farming these small blocks proved not to be economically viable due to unreliable rainfall, damage to crops through pests and disease, production inefficiencies, and marketing difficulties (e.g., Cameron 2005). Close land settlement policies nonetheless prevailed until the 1960s, and continued to test limits to livestock carrying capacity through the calculation of “living areas” based on the economics of grazing enterprises (Casey 1966; Passmore and Brown 1992; Davison and Brodie 2005). The extent of close land settlement under these policies is therefore likely to reflect strong environmental gradients determining livestock carrying capacity (e.g., Hall et al. 1998).
c. Cadastral mapping
In constructing our QGLRI, we exploit the fact that, as a legacy of previous government policies promoting close land settlement, original “grazing selections” remain on Queensland’s cadastral database. Together, these small land parcels span a broad region that, given the small area of the individual land parcels, appears as a “shaded” region on a statewide cadastral map (Fig. 3a). When first gazetted, each grazing selection represented a family property whereas today, as a result of property amalgamation, a family property may span several of these original land parcels. Although Queensland is less closely settled now than it was when families attempted to gain a living from individual grazing selections, the extent of these small grazing selections nonetheless provides an objective basis for defining the geographic limits of more intensive grazing in the past. The western and northern extent of this region can be readily delineated (Fig. 3b) and, as we have argued above, is likely to reflect strong environmental gradients relating to livestock carrying capacity. So, whereas Queensland’s state boundaries better demarcate the actual extent of livestock grazing across the state, the “region of previous close land settlement” (Fig. 3b) better represents where most of that grazing has occurred in the past.
d. Ongoing importance of the region of previous close land settlement for cattle and sheep production
Conversion of sheep grazing enterprises to beef cattle production since the 1990s raises a question as to whether the region of previous close land settlement remains a relevant indicator of livestock distribution patterns in Queensland today. During the period when close land settlement policies prevailed in Queensland (i.e., from the end of World War I until the late 1960s; Cameron 2005), sheep and wool production was the major pastoral industry in western Queensland. However, unfavorable economics since the 1990s has led to a major decline in sheep and wool production across Australia, which, in western Queensland, has involved an ongoing conversion of grazing enterprises to beef cattle production. As such, the northern and western boundaries of the region of previous close land settlement (Fig. 3b) are therefore more likely to reflect environmental constraints to sheep grazing in the past, rather than current limits to cattle grazing. Hall et al. (1998) argue that sheep grazing in far-northern Queensland is limited by a high risk of flooding and the unsuitability of tropical tall grass pastures and, in far-western Queensland, by the severity of the environment (i.e., highly variable and, on average, low pasture growth) together with predation by wild dogs. Grazing by beef cattle is less constrained by such factors.
To assess recent livestock density patterns, we compare the region of previous close land settlement with an objective estimate of the livestock distribution (beef cattle plus sheep, Fig. 4a; and sheep alone, Fig. 4b) based on the 2000/01 ABS census. Queensland is rarely completely drought free as it was leading up to the 2000/01 census, so the livestock distribution in this census year is most likely to reflect longer-term constraints to grazing, rather than shorter-term constraints associated with regional drought. Figure 4a confirms that the region of previous close land settlement includes the most heavily stocked area of Queensland. A distinct “sheep zone” is apparent in western Queensland (Fig. 4b) and, as we have assumed, the western boundary of the region of previous close land settlement aligns quite closely with the western extent of this sheep zone (Fig. 4b). The relatively low livestock density calculated over the sheep zone is consistent with structural changes (e.g., property amalgamation) aimed at addressing overgrazing in southwestern Queensland (Warrego Graziers Association 1988).
Region of previous close land settlement in Queensland (see Fig. 3) in relation to the density of (a) beef cattle and sheep and (b) sheep alone. Livestock numbers are based on the 2000/01 ABS census and standardized to a 450-kg AE. The units of livestock density is AE per kilometer squared.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Region of previous close land settlement in Queensland (see Fig. 3) in relation to the density of (a) beef cattle and sheep and (b) sheep alone. Livestock numbers are based on the 2000/01 ABS census and standardized to a 450-kg AE. The units of livestock density is AE per kilometer squared.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Region of previous close land settlement in Queensland (see Fig. 3) in relation to the density of (a) beef cattle and sheep and (b) sheep alone. Livestock numbers are based on the 2000/01 ABS census and standardized to a 450-kg AE. The units of livestock density is AE per kilometer squared.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
e. Our Queensland grazing lands region
Having delineated the region of previous close settlement (Fig. 3b), and having established that the majority of livestock grazing in Queensland has occurred, and continues to occur, within this area, we use this region to determine the spatial domain of our rainfall index. For ease of description and data extraction, we use straight lines and coordinates based on whole degrees of latitude and longitude, to approximate the northern, western, and southern boundaries of this region (Fig. 3b), thus forming our QGL region. Our QGL region (Fig. 1) is defined as that part of the Australian mainland from 19° to 29°S and east of a line between coordinates 29°S, 144°E and 19°S, 140°E. As such, the western boundary of our QGL region lies approximately parallel to Queensland’s eastern coastline, some 800–1000 km inland, and extends from near the town of Cloncurry in the north to near the town of Hungerford in the south (Fig. 1). The southern boundary (29°S) of our QGL region corresponds with the western portion of the Queensland–New South Wales border and encroaches slightly into New South Wales in the southeast (Fig. 1). Our QGL region encompasses almost the entire region of previous close land settlement, except for a small coastal and subcoastal portion that lies to the north of 19°S (Fig. 3b). This “excluded” portion (Fig. 3b) is a region within which close land settlement is largely the result of intensive agricultural production, including coastal sugar cane production, grazing by dairy cattle, and the production of a diverse range of horticulture crops, rather than extensive beef cattle production per se.
For comparative purposes, we divide the remainder of Queensland into two regions by extending the northern boundary of our QGL region to the Queensland–Northern Territory border (Fig. 1). The region to the north of this boundary is referred to here as far-northern Queensland’ (FNQ), and the region to the west of our QGL region as far-western Queensland (FWQ). Having defined these three Queensland subregions, we can now quantify the relative importance of our QGL region for livestock production. Table 1 shows the area of each subregion and, based on our analysis of 2000/01 ABS census data, livestock units and their density in each subregion. Our calculations indicate that the QGL region, while covering only 59% of Queensland’s land area, carried 83% of Queensland’s adult beef equivalents (AEs, including both cattle and sheep) in 2000/01. Furthermore, the density of AEs in the QGL region was approximately 3 times that of FNQ and 4 times that of FWQ (Table 1).
Importance of Queensland regions in terms of livestock (cattle and sheep production). Livestock numbers are based on the 2000/01 ABS census (see text) and standardized to a 450-kg AE.
4. Selecting a season for our QGLRI
a. Seasons from a climate science perspective
Although the four European seasons of summer, autumn, winter, and spring are less meaningful in subtropical and tropical Australia than in southern Australia, or indeed Europe, this seasonal classification has nonetheless been commonly applied in the study of Queensland rainfall (e.g., McBride and Nicholls 1983; Risbey et al. 2009; Klingaman et al. 2013). However, given the strong seasonality of rainfall in Queensland, it has also been convenient for climate science to base studies on two, rather than four, equal seasons. The warmer and wetter half of the year is referred to either as “summer” (in the subtropics) or “the wet season” (in the northern tropics) and has been typically defined as a 6-month season arbitrarily spanning either October–March (e.g., Lough 1991, 1997) or November–April (e.g., Dick 1958; Weston 1988). However, the BoM now defines the northern wet season in Australia as a 7-month season spanning October–April (BoM 2010).
b. Seasons from an ecological and land-use perspective
For many thousands of years, aboriginal communities across Australia have, depending on location, recognized as few as two or three, but more typically as many as five to seven, distinct “seasons” (Entwisle 2014). Unlike climate science, which has inherited the Gregorian calendar, traditional owners of the land in Australia have not been constrained by the concept of calendar months. Rather, local indigenous communities across Australia have recognized seasons based on changes in the natural environment, which, although in some cases subtle, are nonetheless meaningful in terms of human activity. Such an approach, if translated to the Gregorian calendar, could potentially lead to a more meaningful definition of seasons in Australia (Entwisle 2014).
Within the context of developing a single rainfall index for the grazing lands in Queensland, the choice of two equal-length seasons, while being arithmetically convenient, does not necessarily take into account the varying effectiveness of rainfall on pasture production throughout the year, or other ecological processes important to grazing production and the sustainability of enterprises. Rather than accept a somewhat arbitrary equal arithmetic breakdown of the seasonal cycle, we follow the lead of traditional owners of the land and, as described below, give precedence to the impacts of rainfall on pasture and livestock production across our QGL region, and on the sustainability of the grazed resource.
c. A 5-month (NDJFM) “summer” season
Average monthly rainfall for our QGL region shows a distinct warm season dominance in rainfall, with the three months from December to February contributing 48% of the annual rainfall (Fig. 5). As a consequence, most of the pastures across the region are dominated by C4 grasses, which are better adapted than C3 grasses and dicots to growing under the higher temperatures of the warm season (Fitzpatrick and Nix 1970). Growth of C4 grasses is restricted when the mean weekly temperature falls below 24°C and is severely restricted when the mean weekly temperature falls below 14°C (McCown 1981a). This temperature constraint provides an overriding climatic boundary to the period when pasture growth is possible (Fig. 5). For much of northern Australia, growth of C4 grasses is restricted by either the temperature or the availability of moisture from May to September (Mott et al. 1985), so the BoM’s definition of a 7-month wet season spanning October–April is consistent with the “pasture growing season” across northern Australia as a whole. However, giving consideration to our QGL region per se, in particular the influence of October and April rainfall on the pasture and animal components of grazing enterprises within this region (below), we have elected not to include either October or April rainfall in our QGLRI. These two months each contribute, on average, 6% of the annual (July–June) rainfall and, as we argue below, the inclusion of either month could either exaggerate or mask the severity of summer drought and, therefore, misrepresent the associated risk of degradation to the land and pasture resource. Rather, we suggest that rainfall over the 5-month (NDJFM) season, which contributes 68% of the annual rainfall (Figs. 5 and 6), has the strongest impact on pasture and livestock production and the accompanying managerial decisions.
Average monthly rainfall (bars) and temperature (dot–dashed curves) for the Queensland subregions shown in Fig. 1. The mean weekly temperature below which C4 grass growth is restricted (24°C; McCown 1981a) is shown as a dashed horizontal line. “Average temperature” is the average of daily minima and maxima (see text). The base period for calculations is July 1890–June 2016.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Average monthly rainfall (bars) and temperature (dot–dashed curves) for the Queensland subregions shown in Fig. 1. The mean weekly temperature below which C4 grass growth is restricted (24°C; McCown 1981a) is shown as a dashed horizontal line. “Average temperature” is the average of daily minima and maxima (see text). The base period for calculations is July 1890–June 2016.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Average monthly rainfall (bars) and temperature (dot–dashed curves) for the Queensland subregions shown in Fig. 1. The mean weekly temperature below which C4 grass growth is restricted (24°C; McCown 1981a) is shown as a dashed horizontal line. “Average temperature” is the average of daily minima and maxima (see text). The base period for calculations is July 1890–June 2016.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Queensland rainfall (mm): (a) average annual (July–June), (b) average summer (NDJFM), and (c) average summer rainfall as a proportion of average annual rainfall. The base period for calculations is July 1890–June 2005.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Queensland rainfall (mm): (a) average annual (July–June), (b) average summer (NDJFM), and (c) average summer rainfall as a proportion of average annual rainfall. The base period for calculations is July 1890–June 2005.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Queensland rainfall (mm): (a) average annual (July–June), (b) average summer (NDJFM), and (c) average summer rainfall as a proportion of average annual rainfall. The base period for calculations is July 1890–June 2005.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
d. Influence of October rainfall on pasture and livestock production
October rainfall tends to be associated with infrequent, high-intensity storm events. Ground cover is generally at its lowest at this time of year, so high-intensity rainfall, coupled with low ground cover, results in high loss of water via runoff. High water loss also occurs between storm events, with low ground cover and hot, dry days leading to high bare soil evaporation rates. Pastures have also been observed to respond slowly to available soil moisture at this time of year, due to lower soil temperatures (Tothill 1969) and/or competition with woody C3 plants for limited moisture (Burrows et al. 1988). As such, October rainfall tends not to result in substantial pasture growth.
Pasture burning, as practiced in wetter coastal and subcoastal areas of Queensland, provides a guide to the start of the main pasture growth season in these parts. Burning is a useful management tool to remove senescent pasture components, stimulate new growth, and control both patch grazing and woody plant regrowth. Given the unreliability of October rainfall, it is recommended that burning of pastures in our QGL region take place in early November, after the first storms (Anderson et al. 1988). As such, October rainfall is not relied upon to make a substantial contribution to the bulk of the overall summer feed.
Cattle gain weight mainly while pastures are actively growing and, outside of the main pasture growth season, cattle usually lose or just maintain their liveweight. Once the pasture growth season commences, and new green pasture material with high protein and digestibility becomes readily available, cattle tend to respond with rapid liveweight gains. This generally does not occur until November. Although any pasture grown in October is likely to be highly nutritious, the relatively low quantity available to livestock means that pasture growth at this time will tend to make a limited contribution to the annual cattle liveweight gain. This view is supported by a simulation study of liveweight gain across northern Australia (McCown 1981b), which found, based on historical climate information, that median dates for the commencement of liveweight gain in central and northern Queensland were mid- to late November.
e. Influence of April rainfall on pasture and livestock production
Various interacting factors may restrict pasture growth in April, including low rainfall and soil moisture, flowering of C4 grasses, low nighttime temperatures, reduced rates of nitrogen mineralization, and limits to which available nitrogen can be diluted in plant tissues. However, dry summers tend to delay flowering of C4 grasses and prolong nitrogen mineralization, creating an opportunity for substantial pasture growth in either April or May. In such cases, rainfall during these months can result in rapid pasture growth, although mainly comprising the growth of low quality stem and seed materials. So, while April rainfall may be very effective in producing pasture growth after dry summers, it is less effective in producing animal weight gains. However, our overriding concern, in such cases, is that including high April rainfall in our QGLRI would mask the impact of the preceding dry conditions on both livestock production and the condition of the pasture and soil resource. In particular, if livestock numbers are maintained through a dry summer (NDJFM), low pasture growth leads to high pasture utilization levels (i.e., amount eaten divided by amount grown). This high grazing pressure, in turn, results in poor livestock performance over the summer and, thus, cumulative damage to the pasture and soil resource.
5. Relationship between our QGLRI and airmass dominance
a. Our QGL region in relation to summer airmass distribution
We have assumed that the boundaries of our region of previous close land settlement (Fig. 3) and, in turn, of our QGL region (Fig. 3b) are associated with sharp climate gradients affecting livestock production. In this regard, we note that the boundaries between our QGL, FNQ, and FWQ regions align with the mean summertime position of the ITCZ and Queensland trough, which, in turn, represent boundaries of distinctly different air masses (Fig. 7). An individual air mass, by definition, is horizontally uniform in terms of temperature, stability, humidity, and probable yield of precipitation (Oliver 1970). Differences between air masses, in terms of these properties, can lead to strong gradients in temperature, humidity, and rainfall across boundary regions such as the ITCZ and Queensland trough.
Distribution and origin of major air masses (Em, pTm, tTm, sTc, and Tc) and the general summer flow regime for Queensland (after Sturman and Tapper 2006). The mean February locations of the ITCZ and Queensland trough are indicated.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Distribution and origin of major air masses (Em, pTm, tTm, sTc, and Tc) and the general summer flow regime for Queensland (after Sturman and Tapper 2006). The mean February locations of the ITCZ and Queensland trough are indicated.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Distribution and origin of major air masses (Em, pTm, tTm, sTc, and Tc) and the general summer flow regime for Queensland (after Sturman and Tapper 2006). The mean February locations of the ITCZ and Queensland trough are indicated.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
To the extent that the northern boundary of our QGL regions aligns with the mean summer location of the ITCZ, and the western boundary with the mean summer location of the Queensland trough, our QGL region represents that part of Queensland dominated by the tropical maritime Pacific (pTm) air mass during summer. Furthermore, as the pTm air mass is associated with trade wind flow, and the southern limit of the trade wind flow during summer is approximately 30°S (Rotschi and Lemasson 1967), we can more broadly equate our QGL region with the spatial extent of the pTm airmass dominance over eastern Australia over the summer. It also follows that our FNQ and FWQ regions represent, respectively, those parts of Queensland dominated over the summer by the equatorial maritime (Em) air mass and the tropical continental (Tc) air mass (Sturman and Tapper 2006). As a result, we note strong differences between these regions in terms of climate variables affecting pasture growth and livestock production (i.e., mean rainfall, vapor pressure deficit, and mean daily temperature), which are consistent with differences in characteristic airmass properties (Table 2).
Area-averaged summer (NDJFM) climate of Queensland subregions shown in Fig. 1 compared with the assumed airmass distribution. Variables shown [rainfall, vapor pressure deficit (VPD), and mean daily temperature] each influence pasture growth. Assumed dominant air masses for each region and their respective characteristics (Sturman and Tapper 2006) are shown in parentheses. Means are calculated over 126 summers (1890/91–2015/16).
b. Seasonal timing of pTm airmass dominance over our QGL region
We now consider whether the timing of the seasonal dominance of the pTm air mass over Queensland aligns with our NDJFM summer season. Oliver (1970) provides a ready means of determining, based on the average monthly temperature and rainfall, whether a specific location is dominated by a maritime or continental air mass in a given month. We extend this analysis to spatially averaged rainfall and temperature for our QGL, FNQ, and FWQ regions (Fig. 8). In so doing, we retain the global airmass designations of Oliver (1970), which differ somewhat from the above local classification of air masses in the Australia–New Zealand region (i.e., Sturman and Tapper 2006). To assist in the interpretation, the correspondence between global and local definitions is shown in Table 3.
Thermohyet diagrams (after Oliver 1970) showing monthly (indicated by the first letter of the month, in calendar order) means of air temperature and rainfall throughout the year plotted against realms of dominant airmass regimes [cT (pink), mT/mE (blue), and mP (green); Table 3). Rainfall and temperature data are area averaged over the Queensland subregions shown in Fig. 1: (a) QGL, (b) FNQ, and (c) FWQ. The base period for calculations is July 1890–June 2016.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Thermohyet diagrams (after Oliver 1970) showing monthly (indicated by the first letter of the month, in calendar order) means of air temperature and rainfall throughout the year plotted against realms of dominant airmass regimes [cT (pink), mT/mE (blue), and mP (green); Table 3). Rainfall and temperature data are area averaged over the Queensland subregions shown in Fig. 1: (a) QGL, (b) FNQ, and (c) FWQ. The base period for calculations is July 1890–June 2016.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Thermohyet diagrams (after Oliver 1970) showing monthly (indicated by the first letter of the month, in calendar order) means of air temperature and rainfall throughout the year plotted against realms of dominant airmass regimes [cT (pink), mT/mE (blue), and mP (green); Table 3). Rainfall and temperature data are area averaged over the Queensland subregions shown in Fig. 1: (a) QGL, (b) FNQ, and (c) FWQ. The base period for calculations is July 1890–June 2016.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Assumed relationship between conventional “global” airmass designations (Oliver 1970) and eight regionally defined air masses affecting Australia and New Zealand (Sturman and Tapper 2006).
Figure 8 shows monthly means of air temperature and rainfall through the year plotted against “realms” of dominant airmass regimes (after Oliver 1970). Strong differences between each of the regional charts (Figs. 8a–c) are consistent with inferred differences in summer airmass dominance between each region (Table 2). As such, we can equate Oliver’s (1970) “mT/mE realm” with pTm airmass dominance in our QGL region (Fig. 8a), and Em airmass dominance in FNQ (Fig. 8b). Based on this translation, the seasonal dominance of the pTm air mass extends from mid-November to mid-March in our QGL region (Fig. 8a), corresponding quite closely with the 5-month (NDJFM) summer season upon which we base our QGLRI.
So, just as our QGL region is consistent with the spatial extent of the pTm dominance over eastern Australia, our 5-month season is consistent with the timing of the seasonal dominance of the pTm air mass. In climatological terms, our QGLRI is therefore far more “uniform” than an equivalent statewide rainfall index. In the following section we directly compare both indices, focusing on differences in rainfall variability and its correlation with “local scale” rainfall across Queensland.
6. Comparison of our QGLRI with statewide rainfall
In developing our QGLRI, our intent was to produce an index that, relative to statewide rainfall, better represented rainfall variability experienced by the majority of Queensland’s grazing industry. In this section we highlight differences between our QGLRI and a similarly constructed statewide rainfall time series. As differences can only be attributed to the inclusion–exclusion of both FNQ and FWQ rainfall, we extend our comparisons (below) to both of these subregions, and begin by assessing the relative contribution of each subregion to the statewide average rainfall.
a. Contribution of subregions to statewide average rainfall
As noted above, any difference between our QGLRI and an equivalent statewide time series can be attributed to the inclusion/exclusion of the FNQ and FWQ subregions. The relative contribution of any one subregion to the statewide total is the product of 1) the proportion of Queensland covered by that subregion (on a grid cell rather than an area basis; see section 2) and 2) the average summer rainfall of the region (Table 4). FNQ and FWQ each represent 20% of the state on a grid-cell basis but, given the much higher average rainfall in FNQ than FWQ (913 mm as compared with 183 mm), FNQ rainfall makes, on average, a much higher contribution to the statewide total (41% and 8%, respectively; Table 4). It follows that our QGLRI makes, on average, only a 51% contribution to the statewide rainfall. As such, the potential for an index based on statewide rainfall to misrepresent rainfall variability in our QGL region is quite high and will depend largely on any difference in rainfall variability between the QGL and FNQ subregions.
Average percentage contribution of regional summer (NDJFM) rainfall to statewide summer (NDJFM) rainfall. The contribution of a given region is the product of the proportion of QLD grid cells in that region and the mean rainfall for that region divided by the statewide average (QLD) summer rainfall. The result is expressed in percentage terms (e.g., the average percentage contribution of our QGLRI is 0.596 × 386 mm/451 mm × 100% = 51%).
b. Differences in rainfall variability between regions
We have constructed summer (NDJFM) rainfall time series for each subregion of Queensland as well as for the combined statewide average rainfall. Descriptive statistics for each time series are presented in Table 5 and the individual time series in Fig. 9. To better highlight differences in variability between each time series, in Fig. 9 we plot the “percentage rainfall anomaly” for each summer, rather than total summer rainfall per se. As is evident in Fig. 9, and supported by the calculated coefficient of variation (CV) of each time series (Table 5), summer rainfall averaged over our QGL region is far less variable (CV = 32.7%) than that averaged over FWQ (CV = 52.5%), but far more variable than that averaged over FNQ (CV = 25.2%). Although differences in variability between our QGLRI and the statewide rainfall time series are less obvious, closer comparison (Fig. 10) shows that rainfall anomalies in our QGL region tend to be more extreme. Consistent with the strong contribution of FNQ rainfall to the statewide total (Table 4), the coefficient of variation of statewide rainfall (28.6%) is somewhat lower than that of our QGLRI (32.6%).
Characteristics of area-averaged time series of summer (NDJFM) rainfall. The base period for calculations is 126 summers from 1890/91 to 2015/16.
Time series of summer (NDJFM) rainfall anomalies, expressed as a percent of the respective long-term means, for regions shown in Fig. 1: (a) Queensland, (b) QGL, (c) FNQ, and (d) FWQ. Dashed vertical lines indicate breaks between successive 30-yr periods for which SOI correlations are calculated (Table 7). The base period for calculations is 126 summers from 1890/91 to 2015/16.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Time series of summer (NDJFM) rainfall anomalies, expressed as a percent of the respective long-term means, for regions shown in Fig. 1: (a) Queensland, (b) QGL, (c) FNQ, and (d) FWQ. Dashed vertical lines indicate breaks between successive 30-yr periods for which SOI correlations are calculated (Table 7). The base period for calculations is 126 summers from 1890/91 to 2015/16.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Time series of summer (NDJFM) rainfall anomalies, expressed as a percent of the respective long-term means, for regions shown in Fig. 1: (a) Queensland, (b) QGL, (c) FNQ, and (d) FWQ. Dashed vertical lines indicate breaks between successive 30-yr periods for which SOI correlations are calculated (Table 7). The base period for calculations is 126 summers from 1890/91 to 2015/16.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
(a) Time series and (b) frequency distribution of the difference between the absolute percentage summer (NDJFM) rainfall anomalies for the QGL and Queensland (QLD) regions shown in Fig. 9. The difference in any one summer is calculated as the absolute percentage anomaly for the QGL region minus the absolute percentage anomaly for the QLD region. As the magnitude but not the sign of an anomaly contributes to the variability of a given time series, the analysis highlights summers making the greatest contribution to differences in variability between the QGL and QLD time series (Fig. 9). As shown, positive values indicate that the percentage anomaly (wet or dry) is more extreme over the QGL region than over the whole state (QLD) and vice versa. Summers with “wet” anomalies in the QGL region are shown in blue, and summers with “dry” anomalies in the QGL region are shown in red.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
(a) Time series and (b) frequency distribution of the difference between the absolute percentage summer (NDJFM) rainfall anomalies for the QGL and Queensland (QLD) regions shown in Fig. 9. The difference in any one summer is calculated as the absolute percentage anomaly for the QGL region minus the absolute percentage anomaly for the QLD region. As the magnitude but not the sign of an anomaly contributes to the variability of a given time series, the analysis highlights summers making the greatest contribution to differences in variability between the QGL and QLD time series (Fig. 9). As shown, positive values indicate that the percentage anomaly (wet or dry) is more extreme over the QGL region than over the whole state (QLD) and vice versa. Summers with “wet” anomalies in the QGL region are shown in blue, and summers with “dry” anomalies in the QGL region are shown in red.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
(a) Time series and (b) frequency distribution of the difference between the absolute percentage summer (NDJFM) rainfall anomalies for the QGL and Queensland (QLD) regions shown in Fig. 9. The difference in any one summer is calculated as the absolute percentage anomaly for the QGL region minus the absolute percentage anomaly for the QLD region. As the magnitude but not the sign of an anomaly contributes to the variability of a given time series, the analysis highlights summers making the greatest contribution to differences in variability between the QGL and QLD time series (Fig. 9). As shown, positive values indicate that the percentage anomaly (wet or dry) is more extreme over the QGL region than over the whole state (QLD) and vice versa. Summers with “wet” anomalies in the QGL region are shown in blue, and summers with “dry” anomalies in the QGL region are shown in red.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
c. Differences in rainfall ranking and classification
Rainfall anomalies are often presented as a percentile rank around a mean of 50 (e.g., Ropelewski and Halpert 1987; Lough 1991, 1997), rather than in absolute or percentage terms as above. Ranked values may also be grouped into decile ranges for presentation purposes, or broader quintile or tercile ranges when assessing seasonal forecasts for example (e.g., Potts et al. 1996). Comparing rankings of our QGLRI and statewide rainfall on a percentile basis (Table 6), we find a difference in ranking of more than 10 percentiles in 29% of years and more than 20 percentiles in 4% of years. In one year (1960/61) we observe a 30-percentile difference in rank. Differences in classification occur in 18% of summers for tercile classifications and 30% of summers for quintile classifications. The high frequency of different ranks suggests that caution would need to be applied in translating findings based on a ranking of statewide rainfall to our QGL region.
Percent of grid cells in each region for which summer (NDJFM) rainfall is highly correlated (r ≥ 0.7) with area-averaged (QLD, QGL) rainfall time series (see Figs. 11a,b). Values are rounded to the nearest 5%.
d. Statewide correlation patterns
As a guide to “local” relevance, we consider the correlation of each area-averaged time series with rainfall calculated for individual 0.05° grid cells across Queensland (Fig. 11). We find area-averaged statewide rainfall (Fig. 11a) to be more weakly correlated with rainfall in each subregion (QGL, FNQ, and FWQ) than an index based on that specific subregion (Figs. 11b–d). Area-averaged rainfall over the QGL region (our QGLRI; Fig. 11b) is strongly correlated (r ≥ 0.7) with grid-cell rainfall across 65% of our QGL region whereas, by the same measure, statewide rainfall (Fig. 11a) is only highly correlated with grid-cell rainfall over 50% of our QGL region (Table 6). Moreover, there are few places within our QGL region where, when compared with area-averaged statewide rainfall, our QGLRI is more weakly correlated with grid-cell rainfall (Figs. 11a,b). Thus, for the vast majority of our QGL region, our QGLRI is clearly “more locally relevant” than an equivalent statewide rainfall index.
Simultaneous correlation between area-averaged summer (NDJFM) rainfall time series for regions shown in Fig. 1 with 0.05° × 0.05° resolution rainfall grid cells across Queensland. The correlation coefficient (r value) is mapped across Queensland for each rainfall time series: (a) QLD, (b) QGL summer rainfall (our QGLRI), (c) FNQ summer rainfall, and (d) FWQ summer rainfall. White lines indicate the boundaries of the three regions shown in Fig. 1 on which area-averaged rainfall time series are based. Contours representing r values of 0.7 or higher are highlighted in yellow. The base period for calculations is 126 summers from 1890/91 to 2015/16.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Simultaneous correlation between area-averaged summer (NDJFM) rainfall time series for regions shown in Fig. 1 with 0.05° × 0.05° resolution rainfall grid cells across Queensland. The correlation coefficient (r value) is mapped across Queensland for each rainfall time series: (a) QLD, (b) QGL summer rainfall (our QGLRI), (c) FNQ summer rainfall, and (d) FWQ summer rainfall. White lines indicate the boundaries of the three regions shown in Fig. 1 on which area-averaged rainfall time series are based. Contours representing r values of 0.7 or higher are highlighted in yellow. The base period for calculations is 126 summers from 1890/91 to 2015/16.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Simultaneous correlation between area-averaged summer (NDJFM) rainfall time series for regions shown in Fig. 1 with 0.05° × 0.05° resolution rainfall grid cells across Queensland. The correlation coefficient (r value) is mapped across Queensland for each rainfall time series: (a) QLD, (b) QGL summer rainfall (our QGLRI), (c) FNQ summer rainfall, and (d) FWQ summer rainfall. White lines indicate the boundaries of the three regions shown in Fig. 1 on which area-averaged rainfall time series are based. Contours representing r values of 0.7 or higher are highlighted in yellow. The base period for calculations is 126 summers from 1890/91 to 2015/16.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
7. Our QGL region in relation to ENSO-rainfall response across Queensland
Our intention in developing our QGLRI was to not only better represent the historically strong rainfall variability faced by the majority of the grazing industry in Queensland, but also to better represent how this rainfall variability relates to ENSO, and other climate drivers. ENSO is a major source of variability in summer rainfall across Queensland, and previous studies have highlighted consistent differences in the seasonal timing, duration, and magnitude of the ENSO-rainfall response across the state (e.g., McBride and Nicholls 1983; Ropelewski and Halpert 1987; Lough 1991). Ropelewski and Halpert (1987) defined five Australian “core regions” in terms of a consistent precipitation response to El Niño events and, on this basis, developed summer rainfall time series for each of these regions. Three of these core regions occur in Queensland, and show some correspondence with our Queensland subregions (Fig. 12). The boundaries between these core regions appear to be somewhat consistent with the boundaries of the seasonally dominant air masses over summer as described by Oliver (1970) and Sturman and Tapper (2006). While Ropelewski and Halpert’s (1987) study was limited to El Niño events, and thus did not consider the opposite (La Niña) extreme of ENSO fluctuations, the authors note that their core regions are, in general, consistent with those defined in the ENSO–Australian correlation studies of McBride and Nicholls (1983).
Boundaries between Queensland subregions (straight lines; see Fig. 1) shown in relation to core regions of consistent El Niño–related precipitation response after Ropelewski and Halpert (1987): northern Australia (NAU), central Australia (CAU), and eastern Australia (EAU). Dashed lines indicate areas of possible overlapping boundaries or where boundaries are not well defined.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Boundaries between Queensland subregions (straight lines; see Fig. 1) shown in relation to core regions of consistent El Niño–related precipitation response after Ropelewski and Halpert (1987): northern Australia (NAU), central Australia (CAU), and eastern Australia (EAU). Dashed lines indicate areas of possible overlapping boundaries or where boundaries are not well defined.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
Boundaries between Queensland subregions (straight lines; see Fig. 1) shown in relation to core regions of consistent El Niño–related precipitation response after Ropelewski and Halpert (1987): northern Australia (NAU), central Australia (CAU), and eastern Australia (EAU). Dashed lines indicate areas of possible overlapping boundaries or where boundaries are not well defined.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0148.1
While it is beyond the scope of the current study to undertake a detailed assessment of the relationship between ENSO and Queensland rainfall, including potential links to airmass distribution, it is nonetheless instructive to confirm, as the above studies suggest, that our Queensland subregions (Fig. 1) do indeed differ in terms of the strength of their summer rainfall responses to ENSO fluctuations. As a guide to these responses, we calculate the simultaneous correlation between the SOI and area-averaged summer (NDJFM) rainfall for each of the three regions. Following the approach of Lough (1991), we make this calculation both over the entire record (in this case 126 consecutive summers from 1890/91 to 2015/16), and over consecutive independent 30-yr periods commencing in 1890/91 (Table 7). As outlined by Lough (1991), these specific 30-yr periods align with known changes in the character of ENSO and its relationship with Australian rainfall.
Simultaneous correlation between the SOI and summer (NDJFM) rainfall, spatially averaged over the Queensland regions shown in Fig. 1. Values shown are the correlation coefficient (r) and coefficient of determination (r2; in parentheses). The 30-yr breakdown (i.e., periods 1–4) is based on Lough (1991).
Over the entire 126-yr record we observe a marked difference in the strength of the correlation between the SOI and rainfall for each subregion (Table 7); whereas the SOI accounts for 25% of the variability in QGL rainfall (our QGLRI), the index accounts for a much higher percentage (36%) of the variability in FNQ rainfall and a much lower percentage (13%) of the variability in FWQ rainfall. Moreover, associated with these differences we note a difference in the stationarity of each relationship over independent 30-yr periods (Table 7). FNQ summer rainfall, for example, is not only most strongly correlated with the SOI over the 126-yr period, but also most consistently correlated with the SOI over that period. In contrast, QGL summer rainfall is uncorrelated with the SOI during period 2, and FWQ summer rainfall is uncorrelated with the SOI during both periods 2 and 4. Given the intended use of our rainfall index (section 1), these differences, together with findings from the abovementioned studies, support our constructing a separate rainfall time series based on our QGL region per se.
8. Summary and conclusions
Our QGLRI indicates the area-averaged summer (NDJFM) rainfall over our QGL region (Fig. 1), which, as we have shown in section 3, represents only 60% of Queensland’s land area but carries over 80% of the state’s livestock. We have argued in section 4 that, across this region as a whole, rainfall from November to March is most important for pasture and livestock production. Having defined our QGL region based on close land settlement, we have divided the remainder of Queensland into two regions: FNQ and FWQ. We have shown, in section 5, that boundaries between our three Queensland subregions (QGL, FNQ, and FWQ) correspond to the SPCZ and ITCZ boundaries and, as such, to boundaries between different air masses. Consequently, as we have shown in section 6, our three regions have markedly different climate and rainfall regimes. Despite FNQ and FWQ having a similar area, we have shown that differences in average rainfall between these regions result in FNQ, with high mean rainfall but low variability, making a disproportionately high (41%) contribution to the statewide summer rainfall total when compared with FWQ (8% contribution). Given the above climatic differences between regions, and the high (49%) contribution of FNQ and FWQ rainfall to the statewide total, it would be problematic to infer properties of our QGLRI from an equivalent statewide index. For example, we have shown in section 6 that statewide summer rainfall is less variable than our QGLRI and less representative of rainfall variability within our QGL region. Finally, in section 7, we have noted that our Queensland subregions show some correspondence with Ropelewski and Halpert’s (1987) “core regions” of consistent El Niño-rainfall response over Australia, which may show wider links to the airmass distribution. We have also confirmed that the strength and stationarity of the simultaneous correlation between summer rainfall and the SOI differs strongly between each of our three Queensland subregions.
Referencing Queensland’s state boundaries and European classifications of seasons has obvious limitations for advancing climate science, and the practical relevance of that science in Australia. Our study suggests that referencing airmass boundaries and knowledge of seasonal airmass dominance provides a closer tie to land use, and is likely to be more widely applicable. For example, the spatial and temporal aspects of our QGLRI, as elucidated in this study, could be refined based on more detailed airmass analysis of the Queensland region and, more specifically, factors influencing the onset, decay, and positioning of the ITCZ and Queensland trough. Such studies may also improve our understanding of potential climate change impacts on livestock production in Queensland. Likewise, similar time series to our QGLRI could be developed for regions in northern Australia, which are seasonally dominated by the Tc and Em air masses, thus making the most of the sparse rainfall network across the remainder of Queensland and northern Australia. In this regard we have noted a correspondence between airmass distribution and core regions of uniform ENSO-rainfall response across Australia, which warrants further investigation.
Acknowledgments
The authors acknowledge the valuable assistance of Department of Environment and Science staff, in particular Tracy Van Bruggen for assistance in preparing figures and text, David Ahrens and Keith Moodie for assistance in preparing figures, Dorine Bruget for provision of AussieGRASS data, and Steve Jeffery for assistance with SILO data as well as helpful comments on the text. The authors also acknowledge insightful reviews from three anonymous reviewers. The QGLRI is available online (https://www.longpaddock.qld.gov.au/spota1/papers-and-presentations/).
REFERENCES
Allan, R. J., 1988: El Niño Southern Oscillation influences in the Australasian region. Prog. Phys. Geogr., 12, 313–348, https://doi.org/10.1177/030913338801200301.
Anderson, A. J., A. J. Pressland, R. L. McLennan, R. L. Clem, and K. G. Rickert, 1988: The role of fire in native pasture management. Native Pastures in Queensland: The Resources and Their Management, M. T. Rutherford, W. H. Burrows, and J. C. Scanlan, Eds., Queensland Dept. of Primary Industries Information Series, QI 87023, Queensland Dept. of Primary Industries, 112–124.
BoM, 2010: Average rainfall October to April (Northern wet season). Bureau of Meteorology, http://www.bom.gov.au/jsp/ncc/climate_averages/rainfall/.
Burrows, W. H., J. C. Scanlan, and E. R. Anderson, 1988: Plant ecological relations in open forests, woodlands and shrublands. Native Pastures in Queensland: The Resources and Their Management, Queensland Dept. of Primary Industries Information Series, QI 87023, Queensland Dept. of Primary Industries, 72–90.
Callaghan, J., and S. B. Power, 2010: Variability and decline in the number of severe tropical cyclones making land-fall over eastern Australia since the late nineteenth century. Climate Dyn., 37, 647–662, https://doi.org/10.1007/s00382-010-0883-2.
Cameron, D., 2005: Closer settlement in Queensland: The rise and decline of the agrarian dream, 1860s-1960s. Struggle Country: The Rural Ideal in Twentieth Century Australia, G. Davison and M. Brodie, Eds., Monash University ePress, 129–151.
Carter, J. O., and Coauthors, 1996: Development of data rasters for model inputs. Vol. 3. Development of a National Drought Alert Strategic Information System, Queensland Dept. of Natural Resources, 51 pp.
Carter, J. O., W. B. Hall, K. D. Brook, G. M. McKeon, K. A. Day, and C. J. Paull, 2003: Australian Grassland and Rangeland Assessment by Spatial Simulation (AussieGRASS). Science for Drought: Proceedings of the National Drought Forum, R. Stone and I. Partridge, Eds., Queensland Dept. of Primary Industries, 152–159.
Casey, R. G., 1966: Australian Father and Son. Collins, 102 pp.
Clemens, S. L., H. L. Berry, B. M. McDermott, and C. M. Harper, 2013: Summer of sorrow: Measuring exposure to and impacts of trauma after Queensland’s natural disasters of 2010-2011. Med. J. Aust., 199, 552–555, https://doi.org/10.5694/mja13.10307.
Crimp, S. J., and K. A. Day, 2003: Evaluation of multi-decadal variability in rainfall in Queensland using indices of El Niño-Southern Oscillation and inter-decadal variability. Science of Drought: Proc. National Drought Forum, Brisbane, QLD, Australia, Queensland Dept. of Primary Industries, 106–114.
Davison, G., and M. Brodie, 2005: Struggle Country: The Rural Ideal in Twentieth Century Australia. Monash University ePress, 200 pp., http://www.epress.monash.edu/sc/.
Day, K. A., D. G. Ahrens, A. Peacock, K. G. Rickert, and G. M. McKeon, 2000: Climate tools for northern grassy landscapes. Proc. Northern Grassy Landscapes Conf., Katherine, NT, Australia, Northern Territory Dept. of Primary Industry and Fisheries, https://www.longpaddock.qld.gov.au/spota1/climate-tools-paper/.
Day, K. A., K. G. Rickert, and G. M. McKeon, 2005: Monitoring agricultural drought in Australia. Monitoring and Predicting Agricultural Drought—A Global Study, V. K. Boken, A. P. Cracknell, and R. L. Heathcote, Eds., Oxford University Press, 369–385.
Department of Industry, Innovation and Science, 2016: Australia’s major export commodities: Coal. Dept. of Industry, Innovation and Science Fact Sheet 16-60269, 2 pp., https://industry.gov.au/resource/Mining/AustralianMineralCommodities/Documents/Australias-major-export-commodities-coal-fact-sheet.pdf.
Dick, R. S., 1958: Variability of rainfall in Queensland. J. Trop. Geogr., 11, 32–42.
DSITI, 2014a: SILO climate database—Daily and monthly rainfall. Queensland Spatial Catalogue–QSpatial, Dept. of Science, Information Technology and Innovation, accessed 22 February 2018, http://qldspatial.information.qld.gov.au/.
DSITI, 2014b: SILO climate database—Maximum and minimum temperature. Queensland Spatial Catalogue–Qspatial, Dept. of Science, Information Technology and Innovation, accessed 22 February 2018, http://qldspatial.information.qld.gov.au/.
DSITI, 2014c: SILO climate database—Vapour pressure. Queensland Spatial Catalogue–QSpatial, Dept. of Science, Information Technology and Innovation, accessed 22 February 2018, http://qldspatial.information.qld.gov.au/.
Entwisle, T. J., 2014: Sprinter and Sprummer: Australia’s Changing Seasons. CSIRO Publishing, 168 pp.
Fatichi, S., V. Y. Ivanov, and E. Caporali, 2012: Investigating interannual variability of precipitation at the global scale: Is there a connection with seasonality? J. Climate, 25, 5512–5523, https://doi.org/10.1175/JCLI-D-11-00356.1.
Fitzpatrick, E. A., and H. A. Nix, 1970: The climatic factor in Australian grassland ecology. Australian Grasslands, R. M. Moore, Ed., ANU Press, 1–26.
Folland, C. K., J. A. Renwick, M. J. Salinger, and A. B. Mullan, 2002: Relative influences of the Interdecadal Pacific Oscillation and ENSO on the South Pacific convergence zone. Geophys. Res. Lett., 29, 211–214, https://doi.org/10.1029/2001GL014201.
Ginn, G., 2010: Pastoralism, 1860s–1915. Queensland Historical Atlas, Centre for the Government of Queensland, University of Queensland, http://www.qhatlas.com.au/pastoralism-1860s-1915.
Hall, W. B., G. M. McKeon, J. O. Carter, K. A. Day, S. M. Howden, J. C. Scanlan, P. W. Johnston, and W. H. Burrows, 1998: Climate change in Queensland’s grazing lands: II. An assessment of the impact on animal production from native pastures. Rangeland J., 20, 177–205, https://doi.org/10.1071/RJ9980177.
Klingaman, N. P., S. J. Woolnough, and J. Syktus, 2013: On the drivers of inter-annual and decadal rainfall variability in Queensland, Australia. Int. J. Climatol., 33, 2413–2430, https://doi.org/10.1002/joc.3593.
Lloyd, P. L., and W. H. Burrows, 1988: The importance and economic value of native pastures to Queensland. Native Pastures in Queensland: The Resources and Their Management, W. H. Burrows, J. C. Scanlan, and M. T. Rutherford, Eds., Queensland Dept. of Primary Industries Information Series, QI 87023, Queensland Dept. of Primary Industries, 1–12.
Lough, J. M., 1991: Rainfall variations in Queensland, Australia 1891–1986. Int. J. Climatol., 11, 745–768, https://doi.org/10.1002/joc.3370110704.
Lough, J. M., 1997: Regional indices of climate variation: Temperature and rainfall in Queensland, Australia. Int. J. Climatol., 17, 55–66, https://doi.org/10.1002/(SICI)1097-0088(199701)17:1<55::AID-JOC109>3.0.CO;2-Z.
McBride, J. L., and N. Nicholls, 1983: Seasonal relationships between Australian rainfall and the Southern Oscillation. Mon. Wea. Rev., 111, 1998–2004, https://doi.org/10.1175/1520-0493(1983)111<1998:SRBARA>2.0.CO;2.
McCown, R. L., 1981a: The climatic potential for beef cattle production in tropical Australia: Part I—Simulating the annual cycle of liveweight change. Agric. Syst., 6, 303–317, https://doi.org/10.1016/0308-521X(81)90065-2.
McCown, R. L., 1981b: The climatic potential for beef cattle production in tropical Australia: Part III—Variation in the commencement, cessation and duration of the green season. Agric. Syst., 7, 163–178, https://doi.org/10.1016/0308-521X(81)90044-5.
McKeon, G. M., W. B. Hall, S. J. Crimp, S. M. Howden, R. C. Stone, and D. A. Jones, 1998: Climate change in Queensland’s grazing lands. I. Approaches and climatic trends. Rangeland J., 20, 151–176, https://doi.org/10.1071/RJ9980151.
McKeon, G. M., and Coauthors, 2004: Historical degradation episodes in Australia—Global climate and economic forces and their interaction with rangeland grazing systems. Pasture Degradation and Recovery in Australia’s Rangelands, National Remote Sensing Centre, Brisbane, QLD, Australia, 27–86.
Meat and Livestock Australia, 2016: Australia’s beef industry. MLA, North Sydney, NSW, Australia, https://www.mla.com.au/globalassets/mla-corporate/prices–markets/documents/trends–analysis/fast-facts–maps/mla_beef-fast-facts-2016.pdf.
Miles, R. E., Ed., 1988: Submission to the United Graziers Association on the degradation of south west Queensland. Warrego Graziers Association, Charleville, QLD, Australia, 44 pp., http://www.southwestnrm.org.au/ihub/submission-united-graziers-association-degradation-southwest-queensland.
Mott, J. J., J. Williams, M. H. Andrew, and A. N. Gillison, 1985: Australian savanna ecosystems. Ecology and Management of the World's Savannas, J. C. Tothill and J. J. Mott, Eds., Australian Academy of Science, 56–82.
Nicholls, N., 1992: Historical El Niño/Southern Oscillation variability in the Australasian region. El Niño: Historical and Paleoclimatic Aspects of the Southern Oscillation, H. F. Diaz and V. Markgraf, Eds., Cambridge University Press, 151–173.
Oliver, J. E., 1970: A genetic approach to climatic classification. Ann. Assoc. Amer. Geogr., 60, 615–637, https://doi.org/10.1111/j.1467-8306.1970.tb00750.x.
Passmore, J. G. I., and C. G. Brown, 1992: Property size and rangeland degradation in the Queensland mulga rangelands. Rangeland J., 14, 9–25, https://doi.org/10.1071/RJ9920009.
Potts, J. M., C. K. Folland, I. T. Jolliffe, and D. Sexton, 1996: Revised “LEPS” scores for assessing climate model simulations and long-range forecasts. J. Climate, 9, 34–53, https://doi.org/10.1175/1520-0442(1996)009<0034:RSFACM>2.0.CO;2.
Price Waterhouse Coopers, 2011: Economic impact of Queensland’s natural disasters. Price Waterhouse Coopers, 5 pp., http://pandora.nla.gov.au/pan/130001/20111031-1744/www.pwc.com.au/about-us/flood-support/economic-impact/index.html.
Rimmer, T., and A. W. W. Hossack, 1939: Foreshadowing summer rain in Queensland. Dept. of Physics Rep., Vol. 1, No. 3, University of Queensland, 15 pp., https://espace.library.uq.edu.au/view/UQ:303244/QC1_U7_1939_v1no3.pdf.
Risbey, J. S., M. J. Pook, P. C. McIntosh, M. C. Wheeler, and H. H. Hendon, 2009: On the remote drivers of rainfall variability in Australia. Mon. Wea. Rev., 137, 3233–3253, https://doi.org/10.1175/2009MWR2861.1.
Ropelewski, C. F., and M. S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev., 115, 1606–1626, https://doi.org/10.1175/1520-0493(1987)115<1606:GARSPP>2.0.CO;2.
Rotschi, H., and L. Lemasson, 1967: Oceanography of the Coral and Tasman Seas. Oceanogr. Mar. Biol. Annu. Rev., 5, 49–98.
Skerman, P. J., 1970: Queensland showing land use (showing broad vegetation and soil groups and potential for development). Survey Office Map, Dept. of Lands, Brisbane, QLD, Australia.
Stone, G. S., and Coauthors, 2003: Preventing the 9th degradation episode in Australia’s grazing lands. Science of Drought: Proc. National Drought Forum, Brisbane, QLD, Australia, Queensland Dept. of Primary Industries, 62–73.
Stone, G. S., D. Bruget, J. O. Carter, R. C. Hassett, G. M. McKeon, and D. P. Rayner, 2007: Pasture production and condition. State of the Environment Queensland, J. Freeman and W. Webber, Eds., Environmental Protection Agency, 127–139.
Sturman, A., and N. Tapper, 2006: The Weather and Climate of Australia and New Zealand. 2nd ed. Oxford University Press, 541 pp.
Suppiah, R., 1992: The Australian summer monsoon: A review. Prog. Phys. Geogr., 16, 283–318, https://doi.org/10.1177/030913339201600302.
Syktus, J., G. M. McKeon, N. Flood, I. Smith, and L. Goddard, 2003: Evaluation of a dynamical seasonal climate forecast system for Queensland. Science of Drought: Proc. National Drought Forum, Brisbane, QLD, Australia, Queensland Dept. of Primary Industries, 160–173.
Tothill, J. C., 1969: Soil temperatures and seed burial in relation to the performance of Heteropogon contortus and Themeda australis in burnt native woodland pastures in eastern Queensland. Aust. J. Bot., 17, 269–275, https://doi.org/10.1071/BT9690269.
Treloar, H. M., 1934: Foreshadowing of monsoonal rain in northern Australia. Bureau of Meteorology Bull. 18, 29 pp.
Walker, G. T., and E. W. Bliss, 1930: World weather IV: Some applications to seasonal foreshadowing. Mem. R. Meteor. Soc., 3, 81–95.
Weston, E. J., 1988: The Queensland environment. Native Pastures in Queensland: The Resources and Their Management, W. H. Burrows, J. C. Scanlan, and M. T. Rutherford, Eds., Queensland Dept. of Primary Industries Information Series, QI 87023, Queensland Depat. of Primary Industries, 13–20.
White, W. B., G. M. McKeon, and J. I. Syktus, 2003: Australian drought: The interference of multi-spectral global standing modes and travelling waves. Int. J. Climatol., 23, 631–662, https://doi.org/10.1002/joc.895.
