CPClimate of the PastCPClim. Past1814-9332Copernicus PublicationsGöttingen, Germany10.5194/cp-13-1751-2017Multi-century cool- and warm-season rainfall reconstructions for Australia's major climatic regionsFreundMandymfreund@student.unimelb.edu.auhttps://orcid.org/0000-0002-8839-5494HenleyBenjamin J.KarolyDavid J.AllenKathryn J.https://orcid.org/0000-0002-8403-4552BakerPatrick J.School of Earth Sciences, University of Melbourne, Melbourne, 3010, AustraliaARC Centre of Excellence for Climate System Science, University of Melbourne, Melbourne, 3010, AustraliaAustralian-German Climate and Energy College, University of Melbourne, Melbourne, 3010, AustraliaSchool of Ecosystem and Forest Sciences, University of Melbourne, Richmond, Victoria, 3121, AustraliaMandy Freund (mfreund@student.unimelb.edu.au)30November201713121751177028February201717October201717October201710March2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://cp.copernicus.org/articles/13/1751/2017/cp-13-1751-2017.htmlThe full text article is available as a PDF file from https://cp.copernicus.org/articles/13/1751/2017/cp-13-1751-2017.pdf
Australian seasonal rainfall is strongly affected by large-scale
ocean–atmosphere climate influences. In this study, we exploit the links
between these precipitation influences, regional rainfall
variations, and palaeoclimate proxies in the region to reconstruct Australian
regional rainfall between four and eight centuries into the past. We use an
extensive network of palaeoclimate records from the Southern Hemisphere to
reconstruct cool (April–September) and warm (October–March) season rainfall
in eight natural resource management (NRM) regions spanning the Australian
continent. Our bi-seasonal rainfall reconstruction aligns well with
independent early documentary sources and existing reconstructions.
Critically, this reconstruction allows us, for the first time, to place
recent observations at a bi-seasonal temporal resolution into
a pre-instrumental context, across the entire continent of Australia. We find
that recent 30- and 50-year trends towards wetter conditions in tropical
northern Australia are highly unusual in the multi-century context of our
reconstruction. Recent cool-season drying trends in parts of southern
Australia are very unusual, although not unprecedented, across the
multi-century context. We also use our reconstruction to investigate the
spatial and temporal extent of historical drought events. Our reconstruction
reveals that the spatial extent and duration of the Millennium Drought
(1997–2009) appears either very much below average or unprecedented in
southern Australia over at least the last 400 years. Our reconstruction
identifies a number of severe droughts over the past several centuries that
vary widely in their spatial footprint, highlighting the high degree of
diversity in historical droughts across the Australian continent. We document
distinct characteristics of major droughts in terms of their spatial extent,
duration, intensity, and seasonality. Compared to the three largest droughts
in the instrumental period (Federation Drought, 1895–1903; World War II
Drought, 1939–1945; and the Millennium Drought, 1997–2005), we find that
the historically documented Settlement Drought (1790–1793), Sturt's Drought
(1809–1830) and the Goyder Line Drought (1861–1866) actually had more
regionalised patterns and reduced spatial extents. This seasonal rainfall
reconstruction provides a new opportunity to understand Australian rainfall
variability by contextualising severe droughts and recent trends in
Australia.
Introduction
Australia's climate varies between extreme states of severe dry conditions
and devastating wet episodes affecting large areas of the continent (Nicholls
et al., 1997). Shaped by high variability and persistence, floods, heat waves
and droughts, Australia is highly vulnerable to changes in the climate
system. One reason for the diversity in climate states is the influence of,
and interactions among, large-scale ocean–atmosphere modes of variability.
These include the El Niño–Southern Oscillation (ENSO), the Indian Ocean
Dipole (IOD), the Southern Annular Mode (SAM), and atmospheric
characteristics such as the strength and location of the subtropical ridge
(STR) and the presence of atmospheric blocking (BLK). Critically, these
tropical and extra-tropical modes of variability operate at and across
different temporal scales and their individual and interacting influences
have strong – and diverse – seasonal and regional effects on Australia's
climate (Cai et al., 2014; Drosdowsky, 1993; Larsen and Nicholls, 2009; Maher
and Sherwood, 2014; McBride and Nicholls, 1983; Oliveira and Ambrizzi, 2016;
Ummenhofer et al., 2011; Wang and Hendon, 2007; Watterson, 2009, 2011).
Over the 20th century many regions in Australia have experienced prolonged
pluvial and drought periods that are documented in the gridded, instrumental
records starting in 1900. The Federation Drought (1895–1903) was one of the
first multi-year periods of below average rainfall since European
instrumental data collection began in Australia. There were also pronounced
rainfall deficits during the World War II Drought (1939–1945) and the
Millennium Drought (1997–2005), with devastating effects on regional
agriculture and the broader economy (van Dijk et al., 2013).
In addition to these discrete drought events, there have also been a number
of trends observed in Australian rainfall in recent decades. While there has
been a general decrease in rainfall, particularly across southern Australia,
these changes appear to have strong seasonal and regional components. For
example, rainfall has declined in autumn across southern Australia (Larsen
and Nicholls, 2009; McBride and Nicholls, 1983; Murphy and Timbal, 2008;
Timbal et al., 2006), in the southwest during winter (Allan and Haylock,
1993; Cai and Cowan, 2008; Hope et al., 2009) and in southeast Queensland
during summer (Smith, 2004; Speer et al., 2009). At the same time, regions in
the north have received increasing rainfall (Feng et al., 2013; Taschetto and
England, 2009; Wardle, 2004). The Millennium Drought, observed most severely
in southwestern and southeastern Australia, was predominately due to
deficits in cool-season rainfall (Verdon-Kidd and Kiem, 2009).
Given the presence of decadal (or longer) variability in the known climate
drivers, short observational records are unlikely to provide a reliable
estimate of the full extent of natural variability in Australia's climate
system. In building a picture of the future likelihood of observed late 20th
century trends continuing and the underlying likelihood of prolonged drought,
it is essential that we understand the longer-term climatic context and its
sources of variability. Palaeoclimate data can provide a unique window into
long-term rainfall variability and emerging spatial and temporal trends. Such
knowledge has practical applications for water resources management, seasonal
forecasting, future climate predictions and potential to evaluate simulated
past climate variations.
There have been a number of palaeoclimate reconstructions of hydrological
variables in Australia (Allen et al., 2015; Cullen and Grierson, 2008;
Gallant and Gergis, 2011; Gergis et al., 2011; Heinrich et al., 2009; Lough
et al., 2015). Palmer et al. (2015) recently introduced the Australia and New
Zealand Drought Atlas (ANZDA), using the approach developed for Asia (Monsoon
Asia Drought Atlas (MADA) (Cook et al., 2010)), Europe (Old World Drought
Atlas (OWDA) (E. R. Cook et al., 2015)) and North America by Cook
et al. (2010). The ANZDA reconstructs the past 500 years of the Palmer Drought
severity index (PDSI) for a 0.5∘×0.5∘ grid over
eastern Australia and New Zealand using a network of 176 tree-ring records
and one coral record. Each of these reconstructions has advanced our
knowledge of hydroclimatic variability at the bi-seasonal or annual scale for
specific regions of Australia. To date, however, none have performed
bi-seasonal reconstructions for the entire Australian continent.
The network of palaeoclimate proxies in Australia prior to the instrumental
period is much sparser than for other regions such as Eurasia and North
America. However, the strong links between large-scale remote climate drivers
and Australian climate mean that remote proxies can contain a useful climate
signal. Several recent studies have used remote teleconnections and climate
drivers to obtain skilful reconstructions (Palmer et al., 2015; Tozer et al.,
2016; Vance et al., 2015). In this study, we introduce a new method to
reconstruct regional rainfall by systematically relating instrumental
rainfall and proxy information to remote climate influences. We utilise
a more dynamically focused methodology driven by dynamical relationships to
include remote proxies and maximise the skill and widespread utility of our
reconstructions of Australian rainfall.
Overview of Southern Hemisphere multi-proxy network. (a)
Spatial distribution of tree-ring sites in Tasmania (b) Spatial
distribution of 202 individual proxy records by archive type; national
resource management (NRM) clusters shown on Australian continent
corresponding to abbreviations given in Table 1a. (c) Spatial
distribution of tree-ring sites in New Zealand, (d) Record
availability as a function of archive and time period covered, (e)
Temporal resolution of the palaeoclimate records.
Rainfall variations over the Australian continent show a large degree of
spatial coherence at seasonal and longer time steps, due to the relatively
simple terrain geometries and orography. The Climate Change in Australia
report (CSIRO and Bureau of Meteorology, 2015) applied a regionalisation
scheme to define eight natural resource management (NRM) regions with similar
climatic and biophysical features. The NRM clusters and their abbreviations
are listed in Table 1 and shown on the map in Fig. 1. In this study, we use
a diverse network of local and remote palaeoclimate proxies to perform
a reconstruction of cool- and warm-season rainfall in these eight NRM regions
of Australia.
The aims of this study are as follows:
To consolidate relevant hydroclimate-sensitive palaeoclimate records.
To assess the sensitivity of the palaeoclimate records to the influences of large-scale climate influences and test the stationarity
of these relationships.
To exploit the sensitivity of palaeoclimate proxies to large-scale climate influences and develop skilful palaeoclimate
reconstructions of seasonal rainfall in eight NRM regions for several centuries into the past.
To compare the occurrence of wet and dry periods in the past to those in the instrumental period to provide a longer-term context
for recent observed events and trends.
Our study is organised as follows: Sects. 2 and 3 describe the data and our
methods, respectively. Section 4.1 presents a summary of the regional
rainfall signature of modes of variability in the instrumental period.
Section 4.2 presents the results of the reconstruction. In Sect. 4.3, we
present an investigation of the trends, droughts and extreme years in
a multi-centennial context, as well as a comparison to existing
reconstructions. We finish by discussing these results and their broader
implications in Sect. 5.
Summary of climate drivers, regions and droughts used in this study.
(a) Climate indices and references for computational information; (b) mean, minimal and maximal
seasonal contributions to annual rainfall totals (in %) for natural
resource management (NRM) regions of Australia; and (c)
instrumental and historical droughts.
Our analysis is based on the Australian Bureau of Meteorology's gridded
monthly precipitation dataset from the Australian Water Availability Project
(AWAP) (Jones et al., 2009). The monthly AWAP dataset is based on
precipitation anomalies generated from varying number of station observations
using the Barnes successive-correction (Koch et al., 1983) and
a three-dimensional smoothing spline interpolation (Hutchinson, 1995).
Seasonal and regional averages are computed from the gridded observational
dataset at its highest spatial resolution of 0.05∘×0.05∘ for the period 1900–2015. The eight natural resource
management (NRM) regions reflect a broad pattern of large-scale rainfall
variability but may not capture finer-scale patterns.
We also use several climate indices to link climate drivers with Australian
rainfall (Table 1a) that have previously been used to characterise the
relationship between rainfall and large-scale drivers (Risbey et al., 2009).
Computation of the individual climate indices strictly follows the
descriptions in references given in Table 1. The indices describe tropical
influences (El Niño–Southern Oscillation (ENSO) and Indian Ocean Dipole
(IOD/DMI); Saji et al., 1999) as well as extra-tropical drivers (Southern
Annular Mode, SAM), the intensity and position of the subtropical ridge
strength (STRI and STRP, Drosdowsky, 1993) and atmospheric blocking (BLK)
(Pook and Gibson, 1999). ENSO is described by multiple indices, each of
which relates to a different aspect of the coupled ocean–atmosphere mode.
Here we use indices that are related to sea surface temperatures anomalies in
the eastern Pacific (NCT) and the western Pacific (NWP) (Ren and Jin, 2011),
the Southern Oscillation index (SOI) that measures the atmospheric component
of ENSO, and the effects of central Pacific type events denoted by the ENSO
Modoki index (EMI) (Ashok et al., 2007).
Palaeoclimate data
A palaeoclimate network of 185 individual records is compiled for the
Southern Hemisphere (Fig. 1a). The multi-proxy network includes local and
remote sites from a broad area that are related either directly or
teleconnected to Australian climate (Fig. 1a). The majority of the records
used are derived from the underlying network of the recently developed
Australia and New Zealand Summer Drought Atlas (ANZDA) (Palmer et al., 2015)
and the Ocean2k project, which is part of the PAGES (Past Global Changes)
programme (Neukom and Gergis, 2012; Tierney et al., 2015). The entire network
includes 131 tree-ring records from the Australasian Pacific area, 36
coral-derived records from the tropical Pacific and Indian oceans, and five
speleothem-derived records. In addition, 13 records derived from Antarctic
ice cores are included, as they are related to relevant large-scale
high-latitude drivers such as SAM (Tozer et al., 2016; Vance et al., 2015).
All records in the network extend back to at least 1880 CE and the majority
cover the past 250 years. No further data treatment has been applied other
than the removal of non-climatic biological trends in tree-ring records using
the signal-free method that preserves much of the medium-frequency
variability (timescales of decades to a century) (Melvin et al., 2008) (see
Table S1 in the Supplement for references and details). Approximately half the records extend
back before 1600 CE. Twenty records extend back to 1200 CE or earlier
(Fig. 1b). Within this network, 160 proxy records are annually resolved
records and 25 are sub-annually resolved (derived from corals) (Fig. 1c).
Sub-annually resolved records are binned into seasonal averages according to
a warm (October–March) and cool season (April–September). Samples within
the seasonal window (six consecutive months) are averaged onto a regular time
grid of two samples a year, whereas the dating of annually resolved records
follows the original author.
MethodologyReconstruction
The ocean–atmosphere processes that influence Australia's hydroclimate have
distinct, but variable, seasonal and geographical characteristics (Risbey
et al., 2009). In this study, we first consider the relationships between the
selected climate indices and warm (October–March) and cool-season
(April–September) rainfall in each NRM region. The influence of each
driver is determined by linear correlation for the concurrent season only. We
exclude lag relationships between each driver and rainfall, which are
generally weaker (Risbey et al., 2009). Relationships between precipitation
and ocean–atmosphere processes can vary in strength over time (Gallant
et al., 2013). We therefore use moving correlation windows (window length =30years) to assess statistically significant (p<0.1) correlations
for temporal stability. A relationship is considered stable if the
interquartile range of windowed correlations remains of the same sign for the
entire period of overlap between the two data sets. This approach ensures not only that the climate drivers have an approximately time-stable
relationship with rainfall but also allows some degree of variation in the
strength of the teleconnection. The same procedure to test stability was
applied to the relationships between each proxy record (as listed in
Table S1) and each climate index (as listed in Table 1a). Only proxies with
a significant and time-stable relationship with an index were used as
predictors for NRM regions with a time-stable relationship between that same
index and precipitation (Table S2).
We use a nested, composite-plus-scale (CPS) approach (Bradley and Jones,
1993; Tierney et al., 2015) to reconstruct regionally averaged rainfall for
each NRM region. Our CPS approach combines principal components of proxy
records into regional composites based on a weighted averaging procedure. The
weight, w, is determined by (1) the coefficient of determination between
each record and its target during the common period (1900–1984) and (2) the
significance of this relationship, w=r2(1-p), where p denotes the
p values of the correlation, similar to Tierney et al. (2015). The
resulting composite is re-scaled to the mean and standard deviation during the calibration
period. It should be noted that not all of the records strictly follow
a normal distribution (Fig. S7 in the Supplement). Using a nested approach entails the
calculation of multiple reconstructions, with each reconstruction, or nest,
extending further back in time but including fewer proxies as the proxies
successively drop out (Fig. 1b). Nests are spliced together to form
a continuous reconstruction. Nests are spliced together, based on the most
replicated nest (most number of records available at a given time), to form
a continuous reconstruction. This process maximises the length of the final
reconstruction and ensures all proxies meeting the selection criteria are
used at each point in time. The common period of palaeoclimate records and
instrumental data (1900–1984) is used for calibration and verification.
During the common period, 60 % of the data are used for calibration
(equal to 51 contiguous years) and the remaining 40 % (34 years)
are used for verification. We assess the sensitivity of our reconstruction to
different calibration and verification periods by shifting our calibration
window of 51 years across the common period in steps of
5 years. These different, but not independent, calibration and
verification periods are used to build an ensemble of seven reconstructions
for the warm and cool seasons.
Each final regional rainfall reconstruction is evaluated against a set of
skill metrics. The coefficient of determination (R2c), is a measure of
variance explained by the reconstruction in the calibration period, and
R2v is the variance explained in the verification period. Further skill
statistics include the reduction of error (RE) and the coefficient of
efficiency (CE), which both indicate statistical skill by positive values
(Cook et al., 1994). Further analysis is conducted on the skilful portion of
the reconstruction (CE >0). We report our best reconstruction as that
which maximises the time-integrated RE.
Analysis
Linear trends in the warm and cool season from instrumental and reconstructed
precipitation are compared by fitting a linear trend line to the time series
in 30- and 50-year moving windows. All trends are normalised by the maximum
occurrence across each region and presented in histograms. All trend
calculations are based on overlapping moving windows with a 1-year time step.
We investigate multi-year instrumental and historical periods of extremely
low rainfall. During the instrumental period, three major droughts are
assessed. These are the Millennium (1997–2009), World War II (1935–1945)
and Federation droughts (1895–1903). Since European settlement, seven
historical droughts are often reported in historical and documentary records.
These include the Settlement Drought (1790–1793), the Murray Darling Basin
Drought (1797–1805), the Great Drought (1809–1814), Sturt's Drought
(1809–1830), Southeast Australia Drought (1836–1845), the Black Thursday
Drought (1849–1866) and the Goyder Line Drought (1861–1866) (Helman, 2009).
Details are provided in Table 1c.
We evaluate historical periods of drought by calculating seasonal and annual
rainfall deciles based on the entire length of available data. For
instrumental data, deciles are relative to the 1900–2014 long-term
climatology, while reconstructed rainfall deciles include all positively
verified years. The extended reconstruction includes all verified
reconstructed years up to 1984, extended for the most recent 30 years
of instrumental data up to 2014. The individual duration of deciles therefore
depends on the time interval covered by each individual drought, as given in
Table 1c. It is important to note that deciles deliver quantitative
statements only in conjunction with a baseline period and duration, which may
differ (example of different baseline periods Fig. S6). Resulting deciles are
then categorised into highest on record, very much above average (10), above
average (8–9), average (4–7), below average (2–3), very much below average
(1) and lowest on record.
NRM regions and their dominant climate influences on (a) cool- and
(b) warm-season rainfall. Centre maps show the climate driver with the highest
correlation to seasonal precipitation according to the NRM regions. The
drivers are summarised into four major categories: ENSO, IOD, SAM/BLK and STR
(see Table 1a). Individual correlations between regional rainfall and each
climate driver index are given in surrounding bar plots. Only significant
correlations exceeding the 10 % significance level are shown.
Drought durations are a challenge to compare using a decile-based approach
alone. We therefore apply the concept of “drought–depth–duration” (DDD),
following Fiddes and Timbal (2017) and Timbal and Fawcett (2013) to compare
droughts of different duration. We present the percentage reduction below the
long-term average of dry episodes ranging from 1 to 10 years. Our
seasonal rainfall reconstruction uses the long-term average of the entire
reconstruction as the baseline and presents the drought depth duration as the
percentage reduction below this long-term average.
Extreme years. Summary of the 10 driest and wettest seasons for
each NRM region for different baselines. Different baselines refer to the
instrumental period (Instru: 1900–2014) and the extended reconstruction
period (Pre-Instru: 1200–2014). Years highlighted in bold are among the
10 highest/lowest values for the entire reconstruction and instrumental
period and therefore referred as extreme. Note the reconstruction period
starts for verified periods only and differs for regions and seasons.
Ranking the seasonal rainfall totals in ascending order identifies extreme
dry and wet years. The 10 highest and lowest years in the best reconstruction
are reported in Table 2. As above, to provide a long-term context, the 10
driest and wettest years during the extended period (both the instrumental
and the reconstruction periods) are identified.
In addition to pure statistical verification, we compare our results to other
studies that have used historical documentary records and palaeoclimate
archives to describe or reconstruct past hydroclimatic variability. Data from
other studies with bi-seasonal resolution are averaged into the same warm and
cool seasons used as the basis for reconstructions in this study. Annually
resolved data are compared to both seasons. Single location records are
compared to each of our regions. For the ANZDA (Palmer et al., 2015), area
averages of the NRM clusters are extracted for comparison with the NRM
regions.
ResultsInfluence of climate drivers on regional rainfall
The influence of climate drivers on rainfall variability across the NRM
regions is significant and widespread. The influence of ENSO stands out
across the tropical north and subtropical regions for both warm and cool
seasons (Fig. 2a and b). Indeed, ENSO explains the greatest proportion of
low-latitude rainfall variance (up to 44 %) during the peak-intensity
season (warm season) in the Wet Tropics. The dominant effect of ENSO
decreases along a north–south gradient, with multiple-drivers becoming more
important in the south. In southwest and southeast Australia (Flatlands,
Southern Slopes, Murray Basin) the influence of mid-latitude pressure systems
encapsulated by SAM and atmospheric blocking (BLK) increases. This
north–south gradient is stronger during the cool season. In southern
Australia (Southern Slopes and the Murray Basin), where cool-season rainfall
dominates, the strength and positive influence of the subtropical ridge
(STRP) is important. In southeastern Australia, the correlation between
rainfall and the STRP is as strong as r=0.78, highlighting the importance
of the subtropical ridge on rainfall. Although the influence of the STRI
dominates rainfall in these regions in the cool season, SAM, BLK and
ENSO still have significant associations with mid-latitude rainfall
(Fig. 2a). While conditions in the tropical Pacific have a strong influence
on warm-season rainfall across the continent, the conditions in the IOD have mostly cool-season impacts, except in the Wet Tropics. These
results are consistent with previous studies (e.g. Risbey et al., 2009).
Maps of calibration and verification statistics for the NRM regions.
Columns from left to right are variance explained in the calibration period
(R2c), variance explained in the verification period (R2v),
reduction of error (RE) and coefficient of efficiency (CE) statistics for
both the cool season (a) and the warm season (b).
Statistics shown apply to the most replicated nest of the reconstruction.
Numbers shown in each region indicate, for explained variance, the year in
which R2c and R2v reduce to half of their maximum value,
respectively; those for RE and CE indicate the year in which RE and CE remain positive,
allowing for brief periods of negative skill of duration less than 5 years.
Australian regional rainfall reconstructions during the instrumental
period (1900–2015). Reconstructed cool-season (a) and warm-season (b)
rainfall is compared with the instrumental. Shaded in grey are
uncertainty estimates based on the ensemble spread.
The reconstructionReconstruction skill
Our reconstruction captures 30–60 % of seasonal rainfall variability
across the regions. Skill statistics of the reconstruction (Fig. 3) show that
the variance explained during the calibration period (1934–1984) is around
37 % (R2c [0.2–0.5]) for the cool season and 34 % (R2c
[0.2–0.6]) for the warm season (Fig. 3a and b). During the independent
verification period (1900–1933), a slightly larger magnitude of variance is
explained by the reconstruction, with about 46 % (R2v [0.2–0.6])
for the cool season and 48 % (R2v [0.3–0.6]) for the warm season.
These high and stable proportions of variance captured by the reconstructions
are found for both seasons. The RE and CE statistics are positive for all
regions, indicating reconstruction skill for both seasons across Australia
(Fig. 3a and b). Given that our reconstruction is based on a nested approach,
with a varying set of proxies over time, we indicate in Fig. 3 the timeframe
during which the individual reconstructions show reliable skill for each
region and season. The years shown on each region for the R2 panels
(Fig. 3a, b, e, and f) indicate the earliest year in which the reconstruction
exceeds half of its maximum skill; years shown on the RE and CE panels
(Fig. 3c, d, g, and h) indicate the earliest year in which the RE and CE
statistics remain positive, allowing for brief periods of negative skill, of
duration less than 5 years. The year shown on the CE plots indicates
the earliest year for which the reconstruction is considered skilful (i.e.
CE >0).
During the cool season, the Central Slopes, Wet Tropics, and South and
Southwestern Flatlands indicate the longest skilful (CE >0)
reconstructions, extending back to 1200, 1260 and 1366, respectively. The
rainfall reconstruction in the Rangelands is only skilful back to 1811 and
represents the shortest skilful reconstruction during the cool season. Most
of south and southeastern Australia in the cool season can be reconstructed
back to at least 1749 (Southern Slopes).
In the warm season, the Southern and Southwestern Flatlands, Wet Tropics, and
Murray Basin warm-season reconstructions are skilful (CE >0) for the
longest period, extending back to 1200, 1200 and 1234, respectively. The warm-season rainfall reconstruction in the Monsoonal North has the shortest
skilful reconstruction, back to 1707. Warm-season reconstructions for south
and southeastern Australia indicate slightly longer periods of skill than the
cool-season reconstructions.
Reconstruction time series
Seasonal time series of our reconstructed rainfall across Australia show high
rainfall variability over the instrumental period (Fig. 4). Interannual and
decadal-scale rainfall variability is well characterised by the
reconstructions. For example, the pluvial periods in the mid-1950s and late
1970s during the warm season in the eastern regions (top three panels) are
remarkably well reconstructed. Although different calibration and
verification periods indicate some differences at interannual scales (grey
shading), encouragingly, decadal variability is well represented by the
ensemble. In particular, seasons which dominate the total rainfall are well
captured in terms of their amplitude, for example the warm season in the Wet
Tropics and cool season in the Murray Basin.
Australian regional rainfall reconstructions in cool and warm
seasons. Regional reconstructions for warm- (red line) and cool-season
rainfall (blue line) from 1600 to 1984, extended with instrumental data from
1985 to 2015. Note that multiple axes (warm-season rainfall according to left axis,
cool season refers to right axis). Shaded in grey are uncertainty estimates
based on the ensemble spread. Red and blue shaded periods indicate in-phase
and out-of-phase relationships between the seasons based on windowed
correlations (30 years)>0.5.
Decadal variability is evident in all warm- and cool-season reconstructions
(Figs. 5, S2 and S3). At decadal timescales, both the warm and cool seasons
show synchronous decades of enhanced/reduced rainfall across different
regions. For example, the very dry warm seasons (Fig. S3) observed in the
1960s (Central Slopes, Murray Basin, Rangelands, Southern Slopes, Southern
and Southwestern Flatlands, Wet Tropics) are also seen in the 1760s across
similar regions (Central Slopes, Murray Basin, Monsoonal North, Southern
Slopes, Southern and Southwestern Flatlands, Wet Tropics). During the
1740–1750s extreme wet conditions prevailed in southern and southeastern
Australia. Rainfall during those 20 years was mostly above average for
much of eastern Australia (Central Slopes, East Coast, Murray Basin). In the
1970s similar regions saw a decade of higher than normal warm-season rainfall
(Central Slopes, East Coast, Rangelands, Wet Tropics). There seems to be no
general pattern of prolonged decadal drought or pluvial conditions associated
with specific regions in our reconstructions.
The magnitude of the warm-season pluvials during the 1970s and 1740–1750s are
not anomalously high based on the cool-season reconstructions (Fig. S2). Only
East Coast and Rangelands show similarly high rainfall amounts, while other
regions show average or sightly drier conditions (Central Slopes). Most of
the cool-season decadal trends show very distinct regional patterns. Wetter
than normal conditions in the 1870s are only evident in the Central Slopes
and East Coast. Even geographically proximate regions such as the Murray
Darling Basin region and the Southern Slopes show dissimilarities in terms of
decadal-scale variability. Overall, warm-season rainfall seems to show
slightly more concurrent decades across the regions than during the cool
season.
To assess the degree to which the reconstructions are seasonally distinct,
shaded areas in Fig. 5 highlight periods when the warm- and cool-season
reconstructions are significantly correlated (30-year moving window,
abs(correlation)>0.5). There are a small number of periods during
which the cool- and warm-season rainfall are in phase (positively correlated,
shown in red) or out of phase (negatively correlated, shown in blue).
However, for the most part, there is little dependence between the seasonal
reconstructions. Although some regions show periods in which warm- and cool-season reconstructed rainfall are in phase, this feature is also present in
some instances in the instrumental period. For example, the 1720s and 1760s are periods in which East Coast warm- and cool-season rainfall are positively correlated. During these
decades, there is a degree of synchronicity, meaning that reduced/increased
rainfall in the cool season is accompanied by reduced/increased rainfall in
the following warm season. From 1820–1840 cool- and warm-season rainfall in
the Southern and Southwestern Flatlands is anti-correlated, indicating
opposing seasonal rainfall totals. Rainfall in the late 1970s/1980s in the
Murray Basin, Southern Slopes and Wet Tropics are examples of positive
inter-seasonal correlations (Fig. 5c, f and h) in the instrumental period.
Whilst some regions have short periods of synchronicity across seasons, this
is observed across both the reconstruction and instrumental periods and the
general pattern is one of seasonal independence.
Australian rainfall and drought in a multi-century contextContextualising recent rainfall trends
Since the start of instrumental records in Australia, several major droughts
and extended pluvial periods have been observed. Some influences can be
regarded as temporary changes in the mean state due to, for example, natural
decadal climate variability from the Interdecadal Pacific Oscillation (IPO;
Henley et al., 2015, 2017). Other changes appear to be more strongly
related to long-term changes in atmospheric circulation. These changes are
spatially and temporally diverse, and strong interannual variability makes it
hard to distinguish between low-frequency variability from externally forced
long-term changes. Here we use our rainfall reconstructions to place recent
observed trends in a long-term centennial context.
The histograms in Fig. 6 summarise all 30 and 50-year linear trends in the
regional reconstructions and instrumental data over the period 1600–2014.
The distributions of trends distinguish between pre-1970 variability (grey),
trends since 1970 or 1950 (depending on the fitted trend length of
30/50 years; shown in light blue/red for cool/warm season) and the
trend from the most recent 30 or 50 years. The regional
reconstructions show recent tendencies towards drier cool seasons in the
south (Murray Basin, Southern Slopes, Southern and Southwestern Flatlands)
and wetter warm seasons in the north (Monsoonal North, Rangelands, Wet
Tropics). The distribution of historical trends derived from the
reconstructions are of a generally Gaussian distribution. Some trends
starting after 1950/1970, including the most recent trend, are shifted
towards the upper and lower quartile range of the pre-1950/1970 trends.
Trends starting after 1950/1970, including the most recent trend (ending in
2014), appear unusual, but not unprecedented. In recent years, during the warm
season, tropical regions in particular show a strong increase in rainfall.
This is strongest in the Monsoonal North, followed by the Rangelands and Wet
Tropics (Fig. S3). Some subtropical regions show a warm-season decrease in
rainfall, strongest in the Southern Slopes, but again not unprecedented. All
regions, except the Monsoonal North, show a decline in rainfall in the cool
season during the most recent 30- and 50-year periods. This decline is most
pronounced in the Southern Slopes, Southern and Southwestern Flatlands and
the Murray Darling Basin. Cool-season rainfall, which contributes the
majority of subtropical rainfall, has clear negative shifts in southern
Australia, compared to earlier trends. In particular, cool-season rainfall in
the Murray Basin saw declines over the last 30 and 50 years of the order of
90 mm.
Contextualising recent observed trends in regional Australian
rainfall. Left panels show regional rainfall reconstructions since 1600 for
the warm (red) and cool season (blue) with the 10-year low-pass Chebyshev
filtered series shown as a black line. Grey bars along the x axis denote
non-verified periods for each reconstruction. Right panels show histograms of
30- and 50-year regional rainfall trends (mm yr-1). Grey shaded bars
indicate the full range of the trends prior to 1970 (for 30-year periods) and
1950 for 50-year periods. Light red/blue colouring highlights the trends
since 1970 (for 30-year periods) or 1950 for 50-year periods. The dark
coloured bars indicate the trend in the most recent period. Bar heights are
normalised by the maximum occurrence for each region.
Contextualising the spatial extent and intensity<?xmltex \hack{\break}?> of past droughts
Extended periods of low rainfall have different characteristics in their
temporal and spatial structure. The assessment of drought risk depends
critically on the range of estimated natural variability. Here we assess the
severity of major drought episodes using deciles across both instrumental
(1900–2014) and multi-century reconstruction (1600–2014) periods. Our
regional rainfall reconstruction comparisons here extend the time span of the
instrumental record by a factor of 4, which enables us to view droughts
such as the Millennium Drought in a very long-term multi-century context.
Deciles based on different datasets, baselines and durations are shown in
Fig. 7. Comparing the spatial pattern of the reconstructed droughts to the
gridded and NRM region instrumental (AWAP) spatial patterns, our
reconstruction depicts the intensity of the two drought events during the
instrumental period quite well. The gridded and regional representation of
the Millennium Drought is indicated as the lowest on record for parts of the
Southern Slopes and the Murray Basin and very much below average for the East
Coast and the South and Southwestern Australia, in agreement with other
studies (Cai et al., 2014; Gergis et al., 2011; Verdon-Kidd and Kiem, 2009).
In the context of the full reconstruction period (1600–2014), the Millennium
Drought remains the worst drought since 1749 in the Southern Slopes
(observable too in Fig. 5f) and very much below average for East Coast,
Murray Basin, and Southern and Southwestern Flatlands. In line with
probability estimates for southeastern Australia by Gergis et al. (2011) and
B. I. Cook et al. (2015), the 12-year period of the Millennium Drought is
unprecedented for the Southern Slopes region. The rainfall reconstruction of
the Murray Basin reveals that periods in the late 1700s and early 1800s and
at the time of the Federation Drought are of similar or larger reductions in
rainfall over a 12-year period (Fig. 5c). The Federation Drought period is
also apparent in the Southern and Southwestern Flatlands reconstruction,
along with other periods of rainfall reductions the late 1600s (see Fig. 5g).
The World War II Drought and the Federation Drought appear to be of somewhat
similar character during the instrumental period in terms of their area and
intensity (Fig. 7 gridded and NRM regional plots). Considering the last
400 years, the World War II Drought is a period of average rainfall
for all regions except the Murray Basin, Central Slopes and East Coast. In
contrast, the Federation Drought (1895–1903) is much higher in intensity and
spatial coverage. In Fig. 7 we compare the Federation Drought during its
instrumental period and full (including pre-instrumental) period. During the
observational period (the latter part), the Federation Drought shows only
slightly below average conditions. In the multi-century context, the
Federation Drought shows a wider extent. Along the east coast (Central
Slopes, East Coast), central parts (Rangelands, Murray Basin, Southern
Slopes) and north Australia (Monsoonal North, Wet Tropics) the Federation
Drought was very much below average and lowest on record for
Monsoonal North and Murray Basin, respectively.
Historical droughts during the pre-instrumental period (Table 1c, Figs. 7
and S6), as documented mainly in southeastern Australia due to the
concentration of European settlement there, are captured by the
reconstruction. The Goyder Line Drought, Sturt's Drought and the Great
Drought appear to have affected only certain distinct regions. The Settlement
Drought shows regions clearly below average, especially coastal regions and
the Murray Basin, similarly to Palmer et al. (2015) and Gergis et al. (2010).
The representation in terms of affected regions aligns very well with
historical reports (e.g. historical reports by Sturt or the definition of the
Goyder Line). Most of the historical droughts have been below average in
certain regions but none of these droughts appear to exceed the spatial
extent and intensity of the three major instrumental-period droughts.
Annual deciles for major droughts. Plots for (a) Significant
drought periods (according to Table 1c) during the instrumental
period. Rankings of drought intensity are shown for three major instrumental
period droughts. Column 1: AWAP gridded rainfall (1900–2014); column 2: NRM
clusters (1900–2014); column 3: regional reconstructions during instrumental
period (1900–1990); column 4: regional reconstructions during a four-century
period (1600–2014). (b) Rankings of major drought periods during
the reconstruction period (1600–2014).
Contextualising the duration and seasonality<?xmltex \hack{\break}?> of past droughts
We apply the concept of drought–depth–duration (DDD) to our reconstructions
to further assess the duration and intensity of the different droughts (see
Sect. 3.2 for details). The DDD plots provide a method to compare the
temporal structure of drought periods. Additionally, our bi-seasonal
reconstruction resolution provides an opportunity to investigate the seasonal
nature of protracted droughts and therefore provides insight into their
climatic influences and potential causes.
Using the DDD analysis we can categorise droughts into long- and short-term
droughts and distinguish the primary season of the droughts. The Millennium
Drought is among the worst droughts in terms of its duration across several
of the NRM regions in southern and eastern Australia (Fig. 8, dark blue
lines). In particular, during the cool season, the reduction in rainfall
extends over very long periods for central and eastern regions (Central
Slopes, East Coast and Murray Basin). By comparing short periods
(2<years) to longer periods (>6years) the Millennium Drought
is revealed as a persistent drought, with its worse impacts being felt over
the longer timeframe. The Murray Darling Basin Drought is predominately
caused by cool-seasonal rainfall deficiencies over long periods, plus
a slight accumulated rainfall deficit during the warm season in the Southern
Slopes. The World War II Drought had similar cool-season rainfall reductions;
however the warm-season rainfall was also affected. These are similar to the
findings of (Verdon-Kidd and Kiem, 2009). The Murray Basin, the East Coast
and the Wet Tropics, all had strong reductions in both cool- and warm-season
rainfall of similar magnitude, about 70 % below the long-term mean. In
some regions, the Federation Drought was a strong warm-season feature
(Central Slopes) while in the tropical regions (Wet Tropics and Monsoonal
North), rainfall reached remarkably low values. The intensity of the
Federation Drought during the first few years seems to be a result of
re-occurring El Niño episodes at that time and highlights ENSO's
different spatial effects on Australian rainfall (Ummenhofer et al., 2009).
Droughts such as the Murray Basin Drought can be clearly identified as an
extended warm-season drought, not only in the Murray Basin but also along the
East Coast and the Rangelands. Periods of severe long-term rainfall
reductions highlight the spatial complexity of rainfall variability. There
was abnormally low rainfall during the Southeast Australia Drought over an
extended period of up to 10 years (Monsoonal North, Murray Basin,
Southern and Southwestern Flatlands). This reduction was clearly a warm-season feature that was most severe along the East Coast and the Central
Slopes. Sturt's Drought is another example of a spatially distinct localised
warm-season drought in southeastern Australia that was also expressed as
a long cool-season drought in the Rangelands. One of the strongest short-term
drought episodes was the Black Thursday Drought in which warm-season rainfall
was 60 % below the long-term average and cool-season rainfall was
30 % below the long-term mean. These deficits are likely to have
contributed to the severe bushfires in 1851.
Extreme years in a long-term context
Severe short-term reductions of rainfall leading to events like the Black
Thursday bushfires make clear the devastating effects that single extreme
seasons can have. Our seasonally resolved reconstruction provides for the
first time the opportunity to assess not only extreme years, but to assess
individual extreme seasons.
Rainfall drought depth duration percentages across the regions. The
percentage reduction below the long-term average of the driest years of
variable duration (1–10 years) within the selected drought periods,
for the cool season (a) and warm season (b).
Existing annually resolved reconstructions are likely to exhibit a specific
seasonal bias, which may dilute the impacts of bi-seasonal effects across the
year. Seasonal windows of drought indices often exhibit an integrated signal
from multiple months to possibly years (Keyantash and Dracup, 2002). The
rainfall reconstructions presented in this study enable us to identify
temporally finer-scale extreme seasons of above/below rainfall. We identified
the driest and wettest seasons for our regions by selecting the 10 strongest
events using the instrumental (1900–2014) and reconstruction (1600–2014)
periods.
Comparison with published studies. Spatial correlation plot of
regional NRM rainfall reconstruction and published studies on drought and
rainfall. Spatial maps show region-wise correlation with the summer Australia
New Zealand Drought Atlas (ANZDA) prior to the instrumental period (1600–1899)
(Palmer et al., 2015), season-wise correlation with high-quality
observational data from southeast Australia (Ashcroft et al., 2014) from
1832 to 1859 and a quantitative/visual comparison with pre-instrumental
documentary sources compiled by (Fenby and Gergis, 2012) for southeast
Australia (Central Slopes reconstruction from this study).
Extreme years identified during the instrumental period reveal regional
dependencies and differentiated seasonal aspects of dry and wet years
(Table 2). In the pre-instrumental period similar patterns of spatially
widespread extreme conditions in multiple regions occur (e.g. extreme dry
1481, 1607, 1760 and 1817 (cool season), extreme wet 1759, 1826, 1871 and
1879 (cool season)). Conditions affecting multiple regions with similar
magnitudes were rarer for wet than dry extremes. Warm-season rainfall during
the first half of the 18th century (1720, 1731, 1732, 1740, 1752) was wettest
across multiple regions, while dry conditions of similar magnitude occurred
most frequently during the latter half of the 18th century (1761, 1760,
1814),
affecting the East Coast, Southern Slopes and the Southwestern Flatlands.
The intensity of widespread extreme conditions such as 2010/2011 (wet) or
1982/1983 (dry) is much reduced during the pre-instrumental period. Based on
our reconstructions pre-instrumental seasons with an amplitude comparable to
the 2010 pluvial include 1759 (East Coast) or 1826 (Central Slopes) and were
only extreme across a few regions. Seasonally explicit extremes highlight the
value of the finer temporal resolution resolved by these reconstructions. In
1833, East Coast reconstructed rainfall shows extreme dry conditions during
the cool season followed by extremely wet conditions in the following warm
season.
Comparing our reconstruction with other studies
Here we compare our results with published studies based on palaeo-records
and early documentary compilations. We begin by comparing our results to the
first spatially resolved reconstruction of Australian and New Zealand summer
drought variability, the ANZDA (Palmer et al., 2015). As there is significant
overlap in proxy records used in this study (∼60 %) and in Palmer
et al. (2015), the two studies are not independent in their source data.
Nevertheless, ANZDA is based on tree-ring records and a single coral record
(which is not seasonally resolved), whereas our reconstruction includes
seasonally resolved corals, as well as speleothem and ice-core records. In
addition, the ANZDA differs substantially in its temporal (DJF) and spatial
resolution, since it is a point-by-point gridded reconstruction, mostly
verified for eastern Australia. The ANZDA also targets the PDSI, which is
based not only on rainfall but also temperature and includes memory effects
that account for soil moisture. The PDSI is therefore more likely to reflect
agricultural droughts than meteorological droughts observed in rainfall.
Figure 9 shows the correlations not accounting for memory effects between the
warm-season (DJF) PDSI reconstruction (ANZDA) with our cool- and warm-season
rainfall reconstructions. Non-significant correlations during the cool season
and highly significant warm-season correlations, of up to 0.67 for the East
Coast region, highlight agreement between our seasonal reconstruction and the
summer season hydroclimatic features detected by the PDSI reconstruction. Correlations with the summer season also reiterate the highly seasonal nature of ANZDA and its bias
(intentional) towards only warm-season drought compared with our
reconstructions. The strong temperature dependence of PDSI may explain some
of the differences inland and why large parts of central Australia and all of
Western Australia were not resolved by the ANZDA reconstruction.
The study by Ashcroft et al. (2014) details early documentary records of fine
temporal resolution across southeastern Australia that are entirely
independent of the data used in this study. The temporal coverage of the
documentary records, however, is often less than 30 years.
Nevertheless, there are strong positive correlations between the
documentary-based record and the rainfall reconstruction for both the warm
and cool seasons over the period 1832–1859 (Fig. 9). There are positive
correlations between these records and our rainfall reconstructions across
large areas in southeastern Australia, especially the Murray Basin (cool
season). Figure 9 also shows a comparison of our reconstructions with
rainfall variability as recorded in 18th–19th century (1788–1860) records
from the populated coastal centres (Ashcroft et al., 2014; Fenby and Gergis,
2012). Years classified by Fenby and Gergis (2012) as dry or wet conditions
are partially reflected in our Central Slopes reconstruction. The years
1790–1793, 1810–1813, and 1836–1837 are consistently classified as dry years,
whereas 1788, 1806 and 1830 coincide with years classified as wet.
Consecutive years of dry conditions from 1820–1828 appear to coincide, but
single years such as 1801 and 1802 appear to be not in agreement with our
reconstruction. This may be the result of more localised rainfall variability
than is resolved by our regional Central Slopes reconstruction. Discrepancies
between our reconstruction and documentary sources might also arise from the
use of annual means, which might dilute a seasonal signal. Comparing annual
means of our seasonal reconstructions (not shown) extreme years such as 1860
(East Coast) stand out and mostly agree with the wet and dry classified years
found by Gergis and Ashcroft (2012) for southeast Australia.
Discussion and conclusions
In this study we used an extensive palaeo-climate proxy network derived from
tree rings, corals, ice cores, and speleothem records to reconstruct
precipitation in the eight NRM regions across Australia. This is the first
Australia-wide reconstruction of seasonal rainfall extending ∼400years into the past. The relationships between climate process
indices were evaluated for the strength and stability of their relationship
with precipitation in each NRM region for both the warm and cool seasons. We
simultaneously assessed the strength and stability of relationships between
individual proxy records and these same processes. This process-based
approach enabled the reconstruction of precipitation based on teleconnections
with major processes known to be related to Australian rainfall (Hendon
et al., 2007; Risbey et al., 2009). The screening of predictors based on the
strength and stability of their relationship with the climate indices
constrains the inclusion of predictors during the instrumental period but
still relies on a continuing stationarity assumption on multidecadal
timescales. All reconstructions successfully verified over the 1749–1984
period, and many back to the early 16th century. A comparison with the ANZDA
reconstruction showed a high level of agreement with eastern Australian
drought conditions during the warm season. Independent high-resolution early
documentary records by Ashcroft et al. (2014) compared well with the cool- and
warm-season reconstructions. It should be pointed out that our seasonal
reconstructions represent rainfall only. Droughts are a result of complex
interactions of various atmospheric variables and interactions at different
timescales. Important factors contributing to droughts are temperature, soil
moisture and evaporation, which are not accounted for in this reconstruction.
We also assessed extreme years of high and low rainfall with published
documentary sources. A majority of years previously identified as having
anomalously high or low rainfall events agree well with our seasonal
reconstructions. Most of those comparisons are confined to southeastern
Australia due to regional biases in documentary records. Nevertheless, these
additional verification approaches highlight the quality of our seasonal
reconstruction and its ability to represent past rainfall variability.
On decadal to multi-decadal timescales, substantial low-frequency
variability is present across the regions and seasons. An investigation of
recent trends revealed evidence for unusual tendencies towards wetter
conditions in the north in the warm season and drier conditions in the south
in the cool season. Northern regions (Monsoonal North, Wet Tropics) have
experienced an increase over the last 30–50 years in rainfall,
predominantly during the warm (wet) season (Nicholls, 2006; Taschetto and
England, 2009), when the majority of rainfall occurs. The significance of this
increase is difficult to determine due to high intrinsic variability in the
tropical north. The extended baseline of our reconstruction helps to place
these changes into a long-term context. In particular, after the 1970s the
increase of warm-season rainfall in the Monsoonal North (and the Rangelands)
is highly unusual relative to past centuries. Possible mechanisms could be
strengthening of the monsoon modulated by the ENSO conditions in the western
Pacific (Fig. 2), intensification and shifts in deep convection related to
the monsoon trough (Taschetto and England, 2009), and anthropogenic forcing
enhancing rainfall and cloud formation (Cai et al., 2014). Possible
enhancement of atmospheric pressure systems over the Indian Ocean in
conjunction with a strengthening of SAM (Feng et al., 2013) is a further
possibility. From the correlation analysis of possible drivers (Fig. 2)
neither the IOD nor SAM are significantly correlated with rainfall in the
northern regions, so this latter possibility seems unlikely.
The tendency towards drier conditions in southern Australia is less clear.
Our analysis showed that the most recent decline in rainfall in the Southern
Slopes is within the range of variability given by the reconstruction. The
cool-season decline in the south is often associated with the intensification
of the subtropical ridge (Timbal and Drosdowsky, 2012; Timbal et al., 2006),
changes in large-scale atmospheric circulation such as the Indian Ocean
(Ummenhofer et al., 2011), and the observed upward trend in SAM (Cai and
Cowan, 2006). All of these processes are significantly linked to interannual
rainfall variability in the south (Fig. 2) but are not particularly unusual
in terms of natural variability resolved by the reconstruction. Strong decadal
to multidecadal variability can be observed across all of the regions that
account for declines similar to the most recently observed trends. At least
in terms of 30- to 50-year trend periods, the declining trends are within the
range of intrinsic reconstructed variability. The declining trend of rainfall
during the cool season is particularly strong in the Murray Basin. The most
recent trends starting after 1950s and 1970s are not unprecedented but are
below the 25th percentile, pointing towards a drying tendency. Further work
could specifically focus on the long-term declining trend for the cool season
in south and southwestern Australia (Fig. 6) observed since around 1910.
In order to assess the severity of droughts in a long-term context,
reconstructions need to be consistent with the representation of droughts in
the instrumental record. All three major droughts during the instrumental
periods are well represented in terms of their spatial extent, intensity and
duration considering the reduced spatial representation of regional averages
(Fig. 7) (cf. Verdon-Kidd and Kiem, 2009). The three protracted droughts
remain severe when considered in the ∼400-year context provided by our
reconstructions, especially in south and southeastern Australia. Although all
three droughts have very distinct spatial footprints, the spatial extent and
concurrent drought conditions across broad areas are quite unique compared to
historical droughts. The Millennium Drought stands out as an unprecedented
drought in the Southern Slopes region (Fig. 7). The World War II Drought seems
not to be as exceptional when considering the full reconstruction period back
to 1600; only central and eastern regions were very much below average. The
Federation Drought, which was mostly prior to the observational period, is
confirmed as one of the worst periods of suppressed rainfall in the Murray
Basin and the Monsoonal North region. Compared to droughts during the longer
period, these three major droughts affected large parts of Australia, whereas
pre-instrumental drought episodes such as the Great Drought appeared to be
much more regionally constrained.
The record breaking wet years 2010–2011 accompanied by the strong La
Niña episode were extreme during the warm season across all regions.
During that time the Murray Darling Basin had the largest rainfall anomalies
since 1900 (CSIRO, 2012). In the much longer context of our rainfall
reconstruction, 2010–2011 was one of the wettest warm seasons in the past
several centuries, not only in Murray Basin but also in Monsoonal North,
Rangelands, Southern Slopes and East Coast. Cook et al. (2016) found similar
results based on the ANZDA reconstruction of spatial variability in the PDSI.
The 1979–1983 period of dry conditions in eastern Australia and the 1983
warm season and 1984 cool-season wet anomalies in the East Coast, Wet Tropics
and Monsoonal North are also captured in the reconstructions. The consecutive
1927 (cool), 1928 (warm) and 1928 (cool) seasons experienced extremely dry
conditions in the Central Slopes region. Interestingly, extreme cool- and warm-season wet conditions affected multiple regions, while extreme dry conditions
appear to have occurred more often in single regions. This relationship
persists into the pre-instrumental period. The known asymmetry of the
different ENSO phases impacting Australian rainfall could explain those
differences (Cai et al., 2012). While El Niño-induced rainfall reductions
are not related to the event amplitude, the La Niña phase of ENSO
produces more extreme pluvial events. This is consistent with observations
that La Niña is more strongly teleconnected to rainfall and hence
extreme rainfall events (Cai et al., 2010; King et al., 2013) than El
Niño events are to rainfall deficit.
The spatial extent of various droughts and pluvial episodes may help to
identify the specific drivers behind the individual events. For 20th century
droughts, the interaction of climate modes and anthropogenic warming may have
played a significant role in the drought episodes (Cai et al., 2014).
However, the longer historical perspective provided by our rainfall
reconstructions could help to attribute those factors in further studies by
considering individual climate modes and their interactions.
New insights about drought characteristics are derived from the comparison of
instrumental and historical droughts of different duration. Independent
palaeoclimate reconstructions could help to relate the rainfall variability
back to the climatic drivers causing these diverse drought characteristics in
terms of their spatial and temporal characteristics and could provide important
insights into the climate modes. Different contributing processes are
difficult to examine with only a few events during the instrumental period.
Droughts confined to a specific season can be distinct and classified by
intensity, extent and duration. Again, the three major drought events during
the instrumental period stand out in terms of their intensity and number of
regions affected. The Millennium Drought as a cool-season feature and the
Federation Drought as a warm-season feature clearly highlight the merits of
a seasonally resolved reconstructions (Fig. 8). Another example is the
Southeast Australia Drought, which had persistent dry conditions during
the warm season for up to 10 years.
Natural variability and, in particular, the low-frequency contributions from
decadal modes such as the IPO could contribute to those dry conditions in
various ways. Additional realisations of multi-year droughts can help to
understand physical mechanisms that lead to dry conditions, untangle
dynamical interactions of various climatic modes, and lead to a better
understanding of future drought risks. Inverse approaches and additional
palaeoclimate reconstructions could help to deduce the climatic drivers
causing these diverse drought characteristics and could provide important
insights into the climate modes, their interactions, and influences on
Australian droughts and pluvials. Our multi-century, bi-seasonal and
spatially resolved reconstructions provide new opportunities to study the
dynamics of meteorological droughts across the Australian continent. For
example, do similar settings such as a positive IOD, a positive ENSO, and
a positive IPO lead to similar impacts on Australian droughts? It is
imperative that we answer questions like this due to Australia's significant
vulnerability to prolonged drought episodes. Future work should consider
further subdivision to resolve finer-scale hydroclimatic patterns important
for regional assessments. This could include a subdivision of large regions
such as the Rangelands and regions of complex rainfall regimes like Tasmania
as suggested by the NRM sub-clusters (CSIRO and Bureau of Meteorology, 2015).
Datasets previously published are available from sources
identified in Table S1 in the Supplement. The primary input dataset (NRM
regional means) and the output datasets (reconstructions) are available from
figshare: https://figshare.com/s/a73ff374933b07c7e13c, with data
permanently available at 10.4225/49/59e3ee30cdbbc (Freund, 2017).
The Supplement related to this article is available online at https://doi.org/10.5194/cp-13-1751-2017-supplement.
The authors declare that they have no conflict of
interest.
This article is part of the special issue “Climate of the past
2000 years: regional and trans-regional syntheses”. It is not associated with a
conference.
Acknowledgements
Mandy Freund and David J. Karoly are supported by the Australian Research
Council Centre of Excellence for Climate System Science (CE110001028).
Benjamin J. Henley is supported through an Australian Research Council
Linkage Project (LP150100062) and is an associate investigator of the
Australian Research Council Centre of Excellence for Climate System Science.
Kathryn J. Allen and Patrick J. Baker are supported through an Australian
Research Council Linkage Project (LP12020811). We thank the Bureau of
Meteorology, the Bureau of Rural Sciences and CSIRO for providing the
Australian Water Availability Project data. This is a contribution to the
Past Global Changes (PAGES) 2k Network. PAGES is supported by the US and
Swiss national science foundations. The authors thank Joelle Gergis,
Jonathan Palmer, Ed Cook, Josephine Brown and Ailie Gallant for their
generous advice on this project. Edited by:
Hans Linderholm Reviewed by: Björn E. Gunnarson and two
anonymous referees
References Allan, R. J. and Haylock, M. R.: Circulation features associated with the winter rainfall decrease in Southwestern
Australia, J. Climate, 6, 1356–1367, 1993.Allen, K. J., Lee, G., Ling, F., Allie, S., Willis, M., and Baker, P. J.: Palaeohydrology in climatological context:
developing the case for use of remote predictors in Australian streamflow reconstructions, Appl. Geogr., 64, 132–152,
10.1016/j.apgeog.2015.09.007, 2015.Ashcroft, L., Gergis, J., and Karoly, D. J.: A historical climate dataset for southeastern Australia, 1788–1859,
Geosci. Data J., 1, 158–178, 10.1002/gdj3.19, 2014.Ashok, K., Behera, S. K., Rao, S. A., Weng, H., and Yamagata, T.: El Niño Modoki and its possible
teleconnection, J. Geophys. Res.-Oceans, 112, C11007, 10.1029/2006JC003798, 2007.Bradley, R. S. and Jonest, P. D.: “Little Ice Age” summer temperature variations: their nature and relevance to recent
global warming trends, Holocene, 3, 367–376, 10.1177/095968369300300409, 1993.Cai, W. and Cowan, T.: SAM and regional rainfall in IPCC AR4 models: can anthropogenic forcing account for southwest
Western Australian winter rainfall reduction?, Geophys. Res. Lett., 33, L24708, 10.1029/2006GL028037, 2006.Cai, W. and Cowan, T.: Dynamics of late autumn rainfall reduction over southeastern Australia, Geophys. Res. Lett., 35,
L09708, 10.1029/2008GL033727, 2008.Cai, W., van Rensch, P., Cowan, T., and Sullivan, A.: Asymmetry in ENSO teleconnection with regional rainfall, its
multidecadal variability, and impact, J. Climate, 23, 4944–4955, 10.1175/2010JCLI3501.1, 2010.Cai, W., van Rensch, P., Cowan, T., and Hendon, H. H.: An asymmetry in the IOD and ENSO teleconnection pathway and its
impact on Australian climate, J. Climate, 25, 6318–6329, 10.1175/JCLI-D-11-00501.1, 2012.Cai, W., Purich, A., Cowan, T., van Rensch, P., and Weller, E.: Did climate change-induced rainfall trends contribute to
the Australian Millennium Drought?, J. Climate, 27, 3145–3168, 10.1175/JCLI-D-13-00322.1, 2014.Cook, B. I., Ault, T. R., and Smerdon, J. E.: Unprecedented 21st century drought risk in the American Southwest and
Central Plains, Sci. Adv., 1, e1400082, 10.1126/sciadv.1400082, 2015.Cook, B. I., Palmer, J. G., Cook, E. R., Turney, C. S. M., Allen, K., Fenwick, P., O'Donnell, A., Lough, J. M.,
Grierson, P. F., Ho, M., and Baker, P. J.: The paleoclimate context and future trajectory of extreme summer hydroclimate in eastern
Australia, J. Geophys. Res.-Atmos., 121, 12820–12838, 10.1002/2016JD024892, 2016. Cook, E. R., Briffa, K. R., and Jones, P. D.: Spatial regression methods in dendroclimatology: a review and comparison of
two techniques, Int. J. Climatol., 14, 379–402, 1994.Cook, E. R., Anchukaitis, K. J., Buckley, B. M., D'Arrigo, R. D., Jacoby, G. C., and Wright, W. E.: Asian Monsoon Failure
and Megadrought During the Last Millennium, Science, 328, 486–489, 10.1126/science.1185188, 2010.Cook, E. R., Seager, R., Kushnir, Y., Briffa, K. R., Büntgen, U., Frank, D., Krusic, P. J., Tegel, W., Van der
Schrier, G., Andreu-Hayles, L., Baillie, M., Baittinger, C., Bleicher, N., Bonde, N., Brown, D., Carrer, M., Cooper, R., Čufar, K.,
Dittmar, C., Esper, J., Griggs, C., Gunnarson, B., Günther, B., Gutierrez, E., Haneca, K., Helama, S., Herzig, F., Heussner, K. U.,
Hofmann, J., Janda, P., Kontic, R., Köse, N., Kyncl, T., Levanič, T., Linderholm, H., Manning, S., Melvin, T. M., Miles, D.,
Neuwirth, B., Nicolussi, K., Nola, P., Panayotov, M., Popa, I., Rothe, A., Seftigen, K., Seim, A., Svarva, H., Svoboda, M., Thun, T.,
Timonen, M., Touchan, R., Trotsiuk, V., Trouet, V., Walder, F., Ważny, T., Wilson, R., and Zang, C.: Old World megadroughts and
pluvials during the Common Era, Sci. Adv., 1, e1500561, 10.1126/sciadv.1500561, 2015.CSIRO: Climate and water availability in southeastern Australia: a synthesis
of findings from Phase 2 of the South Eastern Australian Climate Initiative
(SEACI), CSIRO, Canberra, available at:
http://www.seaci.org (last access: 28 February 2017), 2012. CSIRO and Bureau of Meteorology: Climate Change in Australia Information for Australia's Natural Resource Management
Regions: Technical Report, CSIRO and Bureau of Meteorology, Australia, 2015.Cullen, L. E. and Grierson, P. F.: Multi-decadal scale variability in autumn-winter rainfall in south-western Australia
since 1655 AD as reconstructed from tree rings of Callitris columellaris, Clim. Dynam., 33, 433–444,
10.1007/s00382-008-0457-8, 2008. Drosdowsky, W.: An analysis of Australian seasonal rainfall anomalies – 1950–1987. 2. Temporal variability and
teleconnection patterns, Int. J. Climatol., 13, 111–149, 1993.Fenby, C. and Gergis, J.: Rainfall variations in south-eastern Australia part 1: consolidating evidence from
pre-instrumental documentary sources, 1788–1860, Int. J. Climatol., 33, 2956–2972, 10.1002/joc.3640, 2012.Feng, J., Li, J., and Xu, H.: Increased summer rainfall in northwest Australia linked to southern Indian Ocean climate
variability, J. Geophys. Res.-Atmos., 118, 467–480, 10.1029/2012JD018323, 2013.Fiddes, S. and Timbal, B.: Future impacts of climate change on streamflows across Victoria, Australia: making use of
statistical downscaling, Clim. Res., 71, 219–236, 10.3354/cr01447, 2017.Freund, M.: Multi-century cool and warm season rainfall reconstructions for Australia's major climatic regions:
Data and additional information on a seasonal rainfall reconstruction of major NRM regions based on paleoclimate
data, 10.4225/49/59e3ee30cdbbc, 2017.Gallant, A., Phipps, S. J., and Karoly, D. J.: Nonstationary Australasian teleconnections and implications for
paleoclimate reconstructions, J. Climate, 26, 8827–8849, 10.1175/JCLI-D-12-00338.1, 2013.Gallant, A. J. E. and Gergis, J.: An experimental streamflow reconstruction for the River Murray, Australia, 1783–1988,
Water Resour. Res., 47, W00G04, 10.1029/2010WR009832, 2011.Gergis, J. and Ashcroft, L.: Rainfall variations in south-eastern Australia part 2: a comparison of documentary, early
instrumental and palaeoclimate records, 1788–2008, Int. J. Climatol., 33, 2973–2987, 10.1002/joc.3639, 2012.Gergis, J., Garden, D., and Fenby, C.: The influence of climate on the first european settlement of Australia:
a comparison of weather journals, documentary data and palaeoclimate records, 1788–1793, Environ. Hist., 15, 485–507,
10.1093/envhis/emq079, 2010.Gergis, J., Gallant, A. J. E., Braganza, K., Karoly, D. J., Allen, K., Cullen, L., D'Arrigo, R., Goodwin, I.,
Grierson, P., and McGregor, S.: On the long-term context of the 1997–2009 “Big Dry” in South-Eastern Australia: insights from
a 206-year multi-proxy rainfall reconstruction, Clim. Change, 111, 923–944, 10.1007/s10584-011-0263-x, 2011.Heinrich, I., Weidner, K., Helle, G., Vos, H., Lindesay, J., and Banks, J. C. G.: Interdecadal modulation of the
relationship between ENSO, IPO and precipitation: insights from tree rings in Australia, Clim. Dynam., 33, 63–73,
10.1007/s00382-009-0544-5, 2009. Helman, P.: Droughts in the Murray–Darling Basin Since European Settlement, Griffith University, Southport, 2009.Hendon, H. H., Thompson, D. W. J., and Wheeler, M. C.: Australian rainfall and surface temperature variations associated
with the Southern Hemisphere annular mode, J. Climate, 20, 2452–2467, 10.1175/JCLI4134.1, 2007.Henley, B. J., Gergis, J., Karoly, D. J., Power, S., Kennedy, J., and Folland, C. K.: A tripole index for the Interdecadal
Pacific Oscillation, Clim. Dynam., 45, 1–14, 10.1007/s00382-015-2525-1, 2015.Henley, B. J., Meehl, G., Power, S. B., Folland, C. K., King, A. D., Brown, J. N., Karoly, D. J., Delage, F.,
Gallant, A. J. E., Freund, M., and Neukom, R.: Spatial and temporal agreement in climate model simulations of the Interdecadal
Pacific Oscillation, Environ. Res. Lett., 12, 044011,
10.1088/1748-9326/aa5cc8, 2017.Hope, P., Timbal, B., and Fawcett, R.: Associations between rainfall variability in the southwest and southeast of
Australia and their evolution through time, Int. J. Climatol., 30, 1360–1371, 10.1002/joc.1964, 2009.Hutchinson, M. F.: Interpolating mean rainfall using thin plate smoothing splines, Int. J. Geogr. Inf. Syst., 9, 385–403,
10.1080/02693799508902045, 1995. Jones, D. A., Wang, W., and Fawcett, R.: High-quality spatial climate data-sets for Australia, Aust. Meteorol. Ocean., 58,
233–248, 2009. Keyantash, J. and Dracup, J. A.: The quantification of drought: an evaluation of drought indices, B. Am. Meteorol. Soc.,
83, 1167–1180, 2002.King, A. D., Alexander, L. V., and Donat, M. G.: Asymmetry in the response of eastern Australia extreme rainfall to
low-frequency Pacific variability, Geophys. Res. Lett., 40, 2271–2277, 10.1002/grl.50427, 2013. Koch, S. E., Desjardins, M., and Kocin, P. J.: An interactive barnes objective map analysis scheme for use with satellite
and conventional data, J. Clim. Appl. Meteorol., 22, 1487–1503, 1983.Larsen, S. H. and Nicholls, N.: Southern Australian rainfall and the subtropical ridge: variations, interrelationships,
and trends, Geophys. Res. Lett., 36, L08708, 10.1029/2009GL037786, 2009.Lough, J. M., Lewis, S. E., and Cantin, N. E.: Freshwater impacts in the central Great Barrier Reef: 1648–2011, Coral
Reefs, 34, 739–751, 10.1007/s00338-015-1297-8, 2015.Maher, P. and Sherwood, S. C.: Disentangling the multiple sources of large-scale variability in Australian wintertime
precipitation, J. Climate, 27, 6377–6392, 10.1175/JCLI-D-13-00659.1, 2014. Marshall, G.: AMS Journals Online – trends in the Southern Annular Mode from observations and reanalyses, J.
Climate, 16, 4134–4143, 2003. McBride, J. L. and Nicholls, N.: Seasonal relationships between Australian rainfall and the Southern Oscillation,
Mon. Weather Rev., 111, 1998–2004, 1983.Melvin, T. M. and Briffa, K. R.: A signal-free approach to dendroclimatic standardisation, Dendrochronologia, 26, 71–86,
10.1016/j.dendro.2007.12.001, 2008.Murphy, B. F. and Timbal, B.: A review of recent climate variability and climate change in southeastern Australia,
Int. J. Climatol., 28, 859–879, 10.1002/joc.1627, 2008.Neukom, R. and Gergis, J.: Southern Hemisphere high-resolution palaeoclimate records of the last
2000 years,
Holocene, 22, 501–524, 10.1177/0959683611427335, 2012. Nicholls, N.: Detecting and attributing Australian climate change: a review, Aust. Meteorol. Mag., 55, 199–211, 2006.Nicholls, N., Drosdowsky, W., and Lavery, B.: Australian rainfall variability and change, Weather, 52, 66–72,
10.1002/j.1477-8696.1997.tb06274.x, 1997.Oliveira, F. N. M. and Ambrizzi, T.: The effects of ENSO-types and SAM on the large-scale southern blockings,
Int. J. Climatol., 37, 3067–3081, 10.1002/joc.4899, 2016.Palmer, J. G., Cook, E. R., Turney, C. S. M., Allen, K., Fenwick, P., Cook, B. I., O'Donnell, A., Lough, J., Grierson, P.,
and Baker, P.: Drought variability in the eastern Australia and New Zealand summer drought atlas (ANZDA, CE 1500–2012) modulated by
the Interdecadal Pacific Oscillation, Environ. Res. Lett., 10, 1–12, 10.1088/1748-9326/10/12/124002, 2015. Pook, M. and Gibson, T.: Atmospheric blocking and storm tracks during SOP-1
of the FROST Project, Aust. Meteorol. Mag., 1, 51–60, 1999.Ren, H.-L. and Jin, F.-F.: Niño indices for two types of ENSO, Geophys. Res. Lett., 38, L04704, 10.1029/2010GL046031, 2011.Risbey, J. S., Pook, M. J., and McIntosh, P. C.: On the remote drivers of rainfall variability in Australia, Mon. Weather
Rev., 137, 3233–3253, 10.1175/2009MWR2861.1, 2009.Saji, N. H., Goswami, B. N., and Vinayachandran, P. N.: A dipole mode in the tropical Indian Ocean, Nature, 401, 360–363,
10.1038/43854, 1999. Smith, I.: An assessment of recent trends in Australian rainfall, Aust. Meteorol. Mag., 53, 163–173, 2004.Speer, M. S., Leslie, L. M., and Fierro, A. O.: Australian east coast rainfall decline related to large scale climate
drivers, Clim. Dynam., 36, 1419–1429, 10.1007/s00382-009-0726-1, 2009.Taschetto, A. S. and England, M. H.: An analysis of late twentieth century trends in Australian rainfall,
Int. J. Climatol., 29, 791–807, 10.1002/joc.1736, 2009. Tierney, J. E., Abram, N. J., Anchukaitis, K. J., Evans, M. N., Giry, C., Kilbourne, K. H., Saenger, C. P., Wu, H. C., and
Zinke, J.: Tropical sea surface temperatures for the past four centuries reconstructed from coral archives, Paleoceanography,
30, 226–252, 2015.Timbal, B. and Drosdowsky, W.: The relationship between the decline of Southeastern Australian rainfall and the
strengthening of the subtropical ridge, Int. J. Climatol., 33, 1021–1034, 10.1002/joc.3492, 2012.Timbal, B. and Fawcett, R.: A historical perspective on southeastern Australian rainfall since 1865 using the instrumental
record, J. Climate, 26, 1112–1129, 10.1175/JCLI-D-12-00082.1, 2013.
Timbal, B., Arblaster, J. M., and Power, S.: Attribution of the late-twentieth-century rainfall decline in southwest
Australia, J. Climate, 19, 2046–2062, 2006.Tozer, C. R., Vance, T. R., Roberts, J. L., Kiem, A. S., Curran, M. A. J., and Moy, A. D.: An ice core derived 1013-year
catchment-scale annual rainfall reconstruction in subtropical eastern Australia, Hydrol. Earth Syst. Sci., 20, 1703–1717,
10.5194/hess-20-1703-2016, 2016.Ummenhofer, C. C., England, M. H., McIntosh, P. C., Meyers, G. A., Pook, M. J., Risbey, J. S., Gupta, A. S., and
Taschetto, A. S.: What causes southeast Australia's worst droughts?, Geophys. Res. Lett., 36, L04706, 10.1029/2008GL036801,
2009.Ummenhofer, C. C., Gupta, Sen, A., Briggs, P. R., England, M. H., McIntosh, P. C., Meyers, G. A., Pook, M. J.,
Raupach, M. R., and Risbey, J. S.: Indian and Pacific Ocean influences on Southeast Australian drought and soil moisture, J. Climate,
24, 1313–1336, 10.1175/2010JCLI3475.1, 2011.Vance, T. R., Roberts, J. L., Plummer, C. T., Kiem, A. S., and van Ommen, T. D.: Interdecadal Pacific variability and
eastern Australian megadroughts over the last millennium, Geophys. Res. Lett., 42, 129–137, 10.1002/2014GL062447, 2015.van Dijk, A. I. J. M., Beck, H. E., Crosbie, R. S., de Jeu, R. A. M., Liu, Y. Y., Podger, G. M., Timbal, B., and
Viney, N. R.: The Millennium Drought in southeast Australia (2001–2009): natural and human causes and implications for water
resources, ecosystems, economy, and society, Water Resour. Res., 49, 1040–1057, 10.1002/wrcr.20123, 2013.Verdon-Kidd, D. C. and Kiem, A. S.: Nature and causes of protracted droughts in southeast Australia: comparison between
the Federation, WWII, and Big Dry droughts, Geophys. Res. Lett., 36, L22707, 10.1029/2009GL041067, 2009.Wang, G. and Hendon, H. H.: Sensitivity of Australian Rainfall to Inter–El Niño Variations, J. Climate, 20,
4211–4226, 10.1175/JCLI4228.1, 2007.Wardle, R.: Modeled response of the Australian monsoon to changes in land surface temperatures, Geophys. Res. Lett., 31,
L16205, 10.1029/2004GL020157, 2004.Watterson, I. G.: Components of precipitation and temperature anomalies and change associated with modes of the Southern
Hemisphere, Int. J. Climatol., 29, 809–826, 10.1002/joc.1772, 2009.Watterson, I. G.: Understanding and partitioning future climates for Australian regions from CMIP3 using ocean warming
indices, Climatic Change, 111, 903–922, 10.1007/s10584-011-0166-x, 2011.