You are on page 1of 18

Clim Dyn

DOI 10.1007/s00382-013-1974-7

Simulation and prediction of blocking in the Australian region


and its influence on intra-seasonal rainfall in POAMA-2
A. G. Marshall D. Hudson H. H. Hendon
M. J. Pook O. Alves M. C. Wheeler

Received: 11 February 2013 / Accepted: 12 October 2013


Springer-Verlag Berlin Heidelberg 2013

Abstract We assess the depiction and prediction of


blocking at 140E and its impact on Australian intra-seasonal climate variability in the Bureau of Meteorologys
dynamical intra-seasonal/seasonal forecast model Predictive Ocean Atmosphere Model for Australia version 2
(POAMA-2). The model simulates well the strong seasonality of blocking but underestimates its strength and
frequency increasingly with lead time, particularly after the
first fortnight of the hindcast, in connection with the
models drifting basic state. POAMA-2 reproduces well the
large-scale structure of weekly-mean blocking anomalies
and associated rainfall anomalies over Australia; the
depiction of total blocking in POAMA-2 may be improved
with the reduction of biases in the distribution of Indian
Ocean rainfall via a tropical-extratropical wave teleconnection linking blocking activity at 140E with tropical
variability near Indonesia. POAMA-2 demonstrates the
ability to skilfully predict the daily blocking index out to
16 days lead time for the ensemble mean hindcast, surpassing the average predictive skill of the individual
hindcast members (5 days), the skill obtained from persistence of observed (2 days), and the decorrelation timescale of blocking (3 days). This skilful prediction of the
blocking index, together with effective simulation of
blocking rainfall anomalies, translates into higher skill in
forecasting rainfall in weeks 2 and 3 over much of Australia when blocking is high at the initial time of the
hindcast, compared to when the blocking index is small.
POAMA-2 is thus capable of providing forecast skill for

A. G. Marshall (&)  D. Hudson  H. H. Hendon 


M. J. Pook  O. Alves  M. C. Wheeler
Centre for Australian Weather and Climate Research,
Melbourne/Hobart, Australia
e-mail: Andrew.Marshall@csiro.au

blocking rainfall on the intra-seasonal timescale to meet the


needs of Australian farming communities, whose management practises often rely upon decisions being made a
few weeks ahead.
Keywords Blocking  Prediction  Australia 
Intra-seasonal  POAMA-2  Rainfall

1 Introduction
Atmospheric blocking is recognised as an important climate driver for Australias agricultural districts and water
catchments. Influenced by the configuration of hemispheric
planetary waves, blocking involves large quasi-stationary
anticyclones that form in association with a splitting of the
upper tropospheric westerly airstream into two distinct
branches. This splitting occurs at latitudes that are well
poleward of the sub-tropical high pressure belt (or subtropical ridge). Charney and DeVore (1979) first proposed
that blocking events may arise from topographically-driven
Rossby waves creating drag on the mean westerly flow,
such that a resultant resonant response causes the largescale flow to become locked. Recent modelling studies
support this hypothesis (Zidikheri et al. 2007; OKane et al.
2012; Petoukhov et al. 2013). Southern Hemisphere
blocking patterns are of shorter duration and less intense
than those in the Northern Hemisphere (Coughlan 1983;
Wiedenmann et al. 2002), yet they have a major impact on
mid-latitude weather processes. The Tasman Seasouthwest Pacific sector is the dominant region for blocking to
occur in the Southern Hemisphere, with a maximum frequency of occurrence near south-eastern Australia in winter (Coughlan 1983; Pook and Gibson 1999; McIntosh
et al. 2008). Blocking near Australia also typically occurs

123

A. G. Marshall et al.

south of the Great Australian Bight and over the Indian


Ocean southwest of Western Australia (e.g. Qi et al. 1999).
Although the onset and maintenance of blocking in the
Australian region is not yet fully understood, an expanding
literature base on Australian blocking over the last 15 years
has facilitated much improvement in understanding characteristics of blocking. During periods of blocking,
approaching cold fronts tend to weaken and distort before
being steered south-eastwards, often resulting in extended
dry spells in southern Australia (Pook and Gibson 1999). In
particular, south-western Tasmania experiences a reduction
in orographically-derived rainfall due to the anomalous
absence of westerly and south-westerly wind streams in
association with the blocked flow. As cold fronts are being
steered south-eastwards away from the continent, another
type of significant rain-bearing system is known to affect
southern and inland Australia (Hopkins and Holland 1997;
Qi et al. 1999; Pook et al. 2006). This involves the formation of a relatively small cyclonic component on the
equatorward side of the blocking high which breaks away
to create a cutoff low pressure system (Coughlan 1983) and
acts as the major contributor to rainfall over south-eastern
Australia during the AprilOctober agricultural growing
season (Pook et al. 2006). The rainfall response to blocking
depends upon the precise location of the block; blocking in
the vicinity of the Great Australian Bight favours rainfall
over Western Australia, while blocking further east near
140E favours rainfall over south-eastern Australia (e.g.
Risbey et al. 2009).
Blocking events occur within the weather window
(B10 days) but influence Australian rainfall anomalies
across a range of timescales. The interplay between
atmospheric blocking and other major drivers can greatly
complicate the picture; understanding and predicting these
interactions is one of the great challenges for Australian
weather and climate forecasting. Risbey et al. (2009) have
made good headway in this area by assessing the influence
on Australian monthly rainfall of blocking relative to three
other key climate drivers: the Southern Annular Mode
(SAM), the Indian Ocean Dipole (IOD), and the El NinoSouthern Oscillation (ENSO). Their study provides a
qualitative indication of the dominant driver in each location and season by correlating each driver with rainfall at
every grid point on the Australian continent. Blocking is
the leading correlated driver of rainfall over parts of southeastern Australia in all seasons, with the strongest influence
seen in austral winter and spring (JuneNovember) and the
weakest in summer (December- February). Blocking is not
the only driver at play in this region, however, with the
SAM, IOD and ENSO also having a significant and dominant impact on rainfall over parts of the southeast in most
seasons. Risbey et al. (2009) also show that blocking at
140E is negatively correlated with ENSO, meaning that

123

blocking in the Australian region is more probable during


La Nina and less probable during El Nino. The observational study of Risbey et al. (2009) is clearly an important
step towards better understanding the interaction between
major climate drivers and their impact on rainfall variability in Australia.
Weather and climate models show some notable biases
and deficiencies in simulating the observed location and
frequency of blocking highs and cutoff lows (Katzfey and
McInnes 1996; McIntosh et al. 2008; Pook et al. 2010).
Although reasons for this are not yet well known, model
errors in representing blocking can be linked to the climatological biases of the models (van Ulden and van
Oldenborgh 2006; Scaife et al. 2010; Grose et al. 2012).
The locations of blocks may also be sensitive to regional
gradients of SST and land/sea temperature contrasts (Pook
1994; Scaife et al. 2011). Model horizontal resolution is
important for accurately simulating blocking over the
Euro-Atlantic region (Matsueda et al. 2009), but appears
not to be so for blocking over the North Pacific (Matsueda
et al. 2009) or Australia (Grose et al. 2012). It is also
unclear whether vertical, and in particular stratospheric,
resolution should be important for a models representation
of blocking (Scaife and Knight 2008; Woollings et al.
2010). Interestingly, atmosphere-only climate models with
low tropical variability have been shown to present the
poorest simulation of northern hemisphere blocking
(DAndrea et al. 1998), which may also suggest a largescale blocking teleconnection to the tropical atmosphere.
The predictability of blocking in weather and climate
models is correspondingly problematic; the frequency,
position and intensity of blocking are partly predictable on
synoptic timescales but typically underpredicted at intraseasonal timescales and beyond (e.g. Tibaldi and Molteni
1990; Anderson 1993; Watson and Colucci 2002; McIntosh
et al. 2008). Improved understanding of the observed
dynamics of blocking is paramount in effectively diagnosing model deficiencies that lead to the underrepresentation and underprediction of blocking activity (McIntosh
et al. 2008).
While teleconnections between atmospheric blocking
and Australian climate have been assessed on daily and
monthly/seasonal timescales, little has been done on the
intra-seasonal timescale (beyond 1 week and up to a season). At this timescale there exists a notable gap in climate
prediction capability (e.g. Vitart 2004; Hudson et al.
2011a) and an increasing interest in forecast information
from the Australian farming community (e.g. Meinke and
Stone 2005). Blocking may be important for intra-seasonal
predictions since an increase in predictability at this
timescale may be anticipated in a long-lasting blocking
event. In this paper we provide a comprehensive analysis of
the simulation (Sects. 3 and 4) and prediction (Sect. 5) of

Simulation and prediction of blocking

intra-seasonal climate anomalies over Australia in association with blocking in the Australian region in retrospective forecasts (or hindcasts) using the Bureau of
Meteorologys dynamical seasonal prediction model, the
Predictive Ocean Atmosphere Model for Australia (Alves
et al. 2003) version 2 (POAMA-2). POAMA-2 has been
developed with a focus on generating intra-seasonal forecasts in order to bridge the gap between short-range (e.g.
synoptic) and seasonal forecasting for seamless prediction.
The present study is important for contributing to the
development of operational intra-seasonal forecast products as well as contributing to international research
efforts on improving dynamical prediction on this timescale (e.g. Toth et al. 2007; Gottschalck et al. 2010; Vitart
et al. 2008; Hudson et al. 2011a; Marshall et al. 2012b;
Hudson et al. 2013). Section 2 of this paper provides a
brief description of the POAMA-2 forecast system and the
hindcast and observational/re-analysis datasets. The results
of this work are presented in Sects. 35, and a summary
and concluding remarks are provided in Sect. 6.

2 Data Description
POAMA-2 uses the Bureau of Meteorology unified atmospheric model version 3 (BAM3; Colman et al. 2005;
Wang et al. 2005; Zhong et al. 2006) and the Australian
Community Ocean Model version 2 (ACOM2; Schiller
et al. 1997, 2002; Oke et al. 2005). BAM3 is a spectral
transform model with triangular truncation 47 (T47;
approximately 250 km resolution) and 17 vertical sigma
levels (L17), and includes a mass flux convection scheme
(Tiedke 1989) with a convective available potential energy
relaxation closure (Nordeng 1994). The atmospheric CO2
concentration in the model is fixed at 345 ppm. The land
surface component of BAM3 is a simple bucket model for
soil moisture with a field capacity of 150 mm (Manabe and
Holloway 1975) and has three active soil layers for temperature. ACOM2 is based on the Geophysical Fluid
Dynamics Laboratory Modular Ocean Model version 2
(MOM2) and has a zonal grid spacing of 2 longitude and a
telescoping meridional grid spacing of 0.5 equatorward of
8 latitude, gradually increasing to 1.5 near the poles.
ACOM2 has 25 vertical levels, with 8 levels in the top
120 m, a maximum depth of 5 km and an upper ocean
layer thickness of 15 m. The model uses prescribed sea ice
from a multi-year climatology. BAM3 and ACOM2 are
coupled using the Ocean Atmosphere Sea Ice Soil (OASIS)
coupling software (Valcke et al. 2000) with a coupling
frequency of 3 h.
Forecasts from POAMA-2 use initial conditions provided by separate data assimilation schemes for the ocean,
land and atmosphere components of the global coupled

model. Atmospheric and land initial conditions are provided by the Atmosphere-Land Initialisation scheme (ALI;
Hudson et al. 2011b). ALI creates a set of realistic atmospheric initial states by nudging zonal and meridional
winds, temperatures and humidity from the atmosphere
model of POAMA-2 (run prior to hindcasts being made and
forced with observed SST) toward an observationally based
analysis. ALI nudges to the ECMWF 40-year re-analysis
(ERA-40; Uppala et al. 2005) for the period 1980 to August
2002, and to the Bureau of Meteorologys operational
global numerical weather prediction (NWP) analysis
thereafter. ALI also generates land surface initial conditions that are in balance with the atmospheric condition.
Oceanic initial conditions in POAMA-2 are provided by an
ensemble ocean data assimilation system (PEODAS; Yin
et al. 2011). PEODAS is an approximate form of the
ensemble Kalman filter system and generates an ensemble
of ocean states each day including a central unperturbed
ocean analysis.
Perturbed initial conditions are provided for POAMA-2
forecasts using a coupled breeding technique; these are
required to sample forecast uncertainty due to sensitivity
to initial condition errors. The coupled breeding produces
consistent perturbations to both the ocean and atmosphere
at the initial time of the forecasts. The perturbations are
generated using the coupled oceanatmosphere model and
then rescaled to represent analysis uncertainty, centred
every day and added to the unperturbed analyses. This
new strategy represents a significant milestone in our
development of the POAMA forecast system for intraseasonal prediction; see Hudson et al. (2013) for more
details.
A 33-member ensemble of hindcasts of 9 months
duration is produced starting on the 1st, 11th and 21st day
of each month between 1981 and 2010. POAMA-2 uses a
multi-model approach whereby three model versions each
producing 11 ensemble members are used to create the
33-member ensemble (Hudson et al. 2013). We analyse the
first 8 weeks of daily and weekly-mean zonal wind and
geopotential height at the 500 hPa level (u500 and z500
respectively), zonal wind at 850 hPa (u850), and precipitation (rainfall) for each hindcast. Hindcast anomalies are
formed relative to the hindcast model climatology, which is
a function of both start month and lead time, and thus a
first-order linear correction for model mean bias is made in
a similar fashion to that of Stockdale (1997). We define a
lead time of 1 week as the mean of the first week of each
hindcast.
Finally, we note that the southern-most state of Australia, the island of Tasmania, is not resolved as landmass
in POAMA-2 due to the characteristics of the model grid
and resolution. Stretching across approximately 3 of
longitude and 3 of latitude, Tasmania is dissected by the

123

A. G. Marshall et al.

axis of 4 adjacent model grid boxes (of approximate


resolution 2.5 9 2.5) and thus the model sees the
associated grid points as ocean and does not know about
the land or its complex topography, which gives rise to a
high spatial variation of rainfall across a relatively small
area. Thus we only consider the mainland of Australia for
assessing blocking-related Australian rainfall anomalies in
POAMA-2.
Model generated Australian rainfall data are verified
against the Australian Water Availability Project (AWAP)
gridded rainfall dataset (Jones et al. 2009) interpolated onto
the POAMA-2 spatial grid. Model generated global zonal
wind and geopotential height anomalies are verified against
the global re-analysis 1 dataset from the National Centers
for Environmental Prediction/National Center for Atmospheric Research (NNR1; Kalnay et al. 1996), which spans
the period 19482005 and uses a frozen data assimilationforecast system to provide global, quality-controlled
datasets. The subset of NNR1 output used in this study is
also on a 2.58 global grid for the years 19812010. We
create anomalies, relative to climatology, from the observational/re-analysis datasets for direct comparison with
POAMA-2 anomalies.

3 Simulation of blocking and associated climate


anomalies in POAMA-2
3.1 Blocking index
Atmospheric blocking is represented by a simple index that
describes the degree of splitting of the 500 hPa westerly
airstream. Developed by the Bureau of Meteorology for use
in the Australian region (Pook and Gibson 1999), the index
is defined as:
blocking 0:5U25 U30 U40 2U45 U50 U55 U60
where Uy represents the total 500 hPa zonal wind at latitude yS. We focus on blocking at 140E here, but note that
the performance of POAMA-2 documented in this paper is
consistent for blocking at all longitudes in the Australian
region ranging 120E160E (not shown). 140E is regarded as a typical longitude for blocking in the Australian
region to influence synoptic systems over much of Australia (e.g. Risbey et al. 2009), and in particular over the
Mallee region of south-eastern Australia where approximately 50 % of growing season rainfall is associated with
cutoff lows (Pook et al. 2006). In Fig. 1 (top row) we
present schematic diagrams of typical 500 hPa westerly
wind flow situations for which the blocking index is high
(left) and low (right). A high blocking index describes
enhanced 500 hPa westerly flow at high and low latitudes
with reduced flow at mid-latitudes, whereas a low blocking

123

index describes the opposite situation with enhanced flow


at mid-latitudes.
The observed blocking index is constructed using
NNR1 data. We calculate the weekly blocking index at
140E using weekly-mean u500 data for observations and
for POAMA-2 hindcasts at lead times of 18 weeks to
assess the models depiction of blocking in the Australian
region at intra-seasonal timescales. Composites of the
blocking index magnitude (or strength of blocking) at
140E are shown in Fig. 1 as a function of month of year
(i.e. as an annual climatology) for observations (black
line) and for POAMA-2 hindcasts at lead times of
18 weeks (coloured lines). Model composites are constructed relative to the month that pertains to any given
week, such that, for example, a hindcast initialised at the
beginning of January is included in the January composite
for lead times of 14 weeks and in the February composite for lead times of 58 weeks. The observed blocking index exhibits a strong seasonality, reaching a
maximum in July and a minimum in January. This seasonality is also reflected in the 500 hPa zonal wind mean,
with the strong wintertime blocking signal influencing the
appearance of split flow in the mean westerlies (not
shown). A more detailed description of the seasonal cycle
of blocking in the Australian region is provided by Pook
et al. (2013). POAMA-2 reproduces the seasonality of
blocking reasonably well but under-represents the strength
of the index increasingly with lead time after 2 weeks,
suggesting that the models drifting climatology has an
adverse impact on its depiction of blocking.
We use the weekly blocking index described above for
assessing weekly-mean climate anomalies in the following
sections. Before proceeding, however, we also present in
Fig. 1 the blocking frequency at 140E for each month,
calculated from daily data and defined as the proportion of
days for which the blocking index exceeds a value of zero
for at least 5 consecutive days. This definition requires that
blocking episodes are stationary and persistent, and is
motivated by the use of similar thresholds for calculating
blocking frequency in the northern hemisphere, applied to
other indices (e.g. Tibaldi and Molteni 1990; Scaife et al.
2010). Unsurprisingly, blocking frequency exhibits a
strong seasonality similar to that of the blocking index, and
POAMA-2 also under-represents blocking frequency
increasingly with lead time in association with the models
drifting basic state. This developing error in the blocking
frequency (inferred from the difference between the POAMA-2 curves and that observed) begins to settle down
after about the first month as the forecast progresses. We
investigate the impact of mean state bias in more detail in
Sect. 4. As noted earlier, the under-representation of
blocking in climate models is a common problem currently
facing many modelling groups.

Simulation and prediction of blocking

Fig. 1 Top row presents schematic diagrams of relative 500 hPa


westerly wind flow strengths at various latitudes for which the
blocking index is high (left) and low (right) in the Australian region,
as described in the text. Bottom row shows annual climatologies of the

magnitude (left) and frequency (right) of blocking at 140E for


observations (black line) and for POAMA-2 hindcasts at lead times
ranging 18 weeks (coloured lines, as labelled) over the period
19812010

3.2 Anomalous blocking structure

index anomalies are less than one standard deviation (r)


below the mean, representing a physical situation of
enhanced westerly mid-level flow at mid-latitudes. To
first order, the high and low index polarities have
opposite signed but otherwise identical climate impacts
(not shown), and thus we calculate the high-minus-low
index composite difference to describe anomalous conditions during strong blocking episodes. For both
observations and POAMA-2, 1r & 20, so there are a
few instances in DJF when a high index blocking
anomaly can have a weak negative absolute index.
Exclusion of these cases from the analyses makes no
appreciable difference to our results (not shown). We
perform our assessment for each season separately since
the impact of observed blocking on Australian climate
anomalies varies between austral summer (December
February; DJF), autumn (MarchMay; MAM), winter
(JuneAugust; JJA) and spring (SeptemberNovember;
SON), as shown in Risbey et al. (2009).

To explore the spatial structure of large-scale climate


anomalies associated with blocking and their representation in POAMA-2, we first calculate the weekly
blocking index as an anomaly relative to its annual
climatology, as we are now interested in mean flow
perturbations associated with blocking conditions and
their impact on climate anomalies in the Australian
region at different times of year, irrespective of the
strong seasonality of blocking demonstrated in Fig. 1.
We identify episodes of strong blocking as defined using
high (positive) and low (negative) polarities of the
anomalous blocking index. The high polarity occurs
where blocking index anomalies are greater than one
standard deviation (r) above the mean, representing a
physical situation of the reduced mid-latitude westerly
flow experienced during episodes of strong blocking.
Conversely, the low polarity occurs where blocking

123

A. G. Marshall et al.

Fig. 2 Weekly-mean composites of anomalous 500 hPa geopotential


height for the high-minus-low polarity of blocking at 140E over the
period 19812010 in JJA (top row), SON (second row), DJF (third
row), and MAM (bottom row), for observations (left column) and
POAMA-2 hindcasts averaged over lead times of 36 weeks (right

column). Statistically significant height anomalies are shaded (95 %


confidence level using a t-test applied to the difference of two means).
Contour interval is 20 m with blue shading representing negative
values. The total number of cases for the high and low polarity of the
blocking index is shown in the title of each panel

We assess how effectively POAMA-2 reproduces the


observed broad-scale structure of z500 and u500 anomalies
in association with blocking at 140E. Figures 2 (z500) and
3 (u500) show weekly-mean anomaly composites for
strong blocking episodes over the period 19812010 for
observations and for POAMA-2 hindcasts averaged over
lead times of 36 weeks (noting that the anomalous patterns are similar at these lead times). We use individual
ensemble members for the model analysis, rather than
the ensemble mean, and present statistically significant
anomalies at the 95 % confidence level using a standard
t test applied to the difference of two means (high-minus-

low index). For the model analysis, the number of composite samples N used in the t-test is replaced by Neff

123

N 1q
1q to take account of the fact that ensemble members
may not be independent, although this is more an issue at
shorter lead times, i.e. in the first couple of weeks. The
autocorrelation coefficient q is calculated by correlating
each individual ensemble member with the remaining
ensemble members, at each grid point, as a function of
hindcast start date and lead time. Correlation coefficients
are then averaged over all members to produce a spatial
map of the autocorrelation that is used to calculate Neff at

Simulation and prediction of blocking

Fig. 3 As for Fig. 2 but for anomalous 500 hPa zonal wind in JJA. Contour interval is 2 m/s

each grid point. The number of degrees of freedom is thus


reduced at short lead times when the ensemble spread is
small.
The observed composites in Fig. 2 show positive z500
anomalies centred near 5055S in all seasons, with maxima ranging from about 160 m (SON) to 220 m (MAM), in
association with blocking at 140E. Low (negative) z500
centres appear over southern Australia (near 3035S,
related to the cut-off low component of blocking over
Australia) and south-east of the high (near 6570S and
just east of the dateline). This anomalous z500 structure
appears to be part of a Rossby wave train emanating from a
weak tropical high near Indonesia; an approximate great
circle can be traced through these anomalous z500 centres
together with a much weaker high centred near the South
American continent. (The tropical high in z500 appears
weak because the wave structure is baroclinic near the
equator but translates to barotropic at higher latitudes.
Streamfunction anomalies complement Fig. 2 in clarifying
this tropical link somewhat; not shown.) Further, in addition to the Australian and South American highs, a third
(weak) positive z500 anomaly appears near South Africa
(magnitude 2040 m); together these mid-latitude anomalies resemble the global-scale zonal wave three structure
that often dominates the southern hemisphere circulation
between 45S and 55S (e.g. van Loon and Jenne 1972; Mo
and White 1985; Kidson 1988; Karoly 1989; Raphael
2004) and is known to occur in association with Southern
Hemisphere blocking (van Loon 1956; Trenberth and Mo
1985).
Figure 3 shows the corresponding u500 anomaly for the
Australian region; we present the JJA season only, noting
similarity in the broad-scale patterns for all seasons. The
observed anomalous zonal flow over the southern extremes
of Australia is predominantly easterly with a peak magnitude of around 20 m/s centred near 45S (i.e. between the
positive and negative z500 centres in Fig. 2; co-located

with the maximum z500 meridional gradient). Over the rest


of the Australian continent north of about 33S, the
anomalous zonal flow is primarily westerly with a peak
magnitude ranging 914 m/s centred at 25S; this relates to
the northernmost component of the split westerly flow
associated with blocking. The southernmost component of
the split westerly flow is centred near 6065S with a peak
magnitude ranging 1115 m/s.
POAMA-2 reproduces the observed z500 and u500
blocking anomalies in the Australian region quite well. The
mid-latitude positive z500 anomaly is centred near
5055S in all seasons, consistent with that observed, and
the associated Australian low z500 centre is also centred
near 3035S in the model (Fig. 2). POAMA-2 also
reproduces a similar wave train structure to that observed,
as well as weaker high z500 centres near the South
American and South African continents. There exist some
differences from the observed blocking pattern in the
magnitude of the anomalies: the negative z500 anomaly
centres are weaker than those observed over southern
Australia by about 2035 %, and stronger than those
observed at high latitudes by about 1525 %, in JJA, SON
and DJF. The peak strength of the blocking anticyclone
anomaly is also stronger than that observed in JJA (10 %)
and SON (20 %), however overall the models depiction of
the spatial structure and strength of z500 blocking anomalies is impressive. Correspondingly, POAMA-2 reproduces the blocking split zonal flow characteristics
reasonably well in all seasons (illustrated in Fig. 3 for JJA)
although the discrepancies between the modelled and
observed z500 anomalies translate to the differences in
u500 anomalies seen in Fig. 4. Specifically, the model
experiences stronger anomalous easterly flow over much of
southern Australia, particularly over the southeast, and
stronger anomalous westerly flow in the high latitude split
flow region, with anomaly differences of up to 23 m/s in
all seasons. Thus POAMA-2 generally demonstrates a

123

A. G. Marshall et al.

Fig. 4 POAMA-2 minus observed difference in anomalous 500 hPa


zonal wind based on the weekly-mean composites of Fig. 3 for JJA
(top left), SON (top right), DJF (bottom left) and MAM (bottom

right). Contour interval is 1 m/s with blue shading representing negative values and white shading representing absolute values \1 m/s.
Zero contours are omitted from figure panels

small enhancement in blocking variability relative to that


observed. With this enhanced variability remaining consistent over all lead times (not shown), we find that the
drifting basic state (which affects total blocking; Fig. 1)
has little impact on the strength of blocking anomalies in
POAMA-2; we address this further in Sect. 4.

easterly wind anomalies over the Top End associated


with blocking at 140E (not shown); this is an area for
future examination. The observed weekly rainfall composites of Fig. 5 are broadly consistent with the seasonal
rainfall maps of Fig. 13 in Risbey et al. (2009), which
also show the impact of blocking at 140E. There exist a
few notable differences between the two analyses however, such as reduced areas of statistical significance over
Western Australia in Risbey et al. (2009). Such discrepancies may be explained by some important differences between the methodologies used. For example,
some of the weekly fluctuations in the blocking index
and rainfall are not picked up in the monthly-mean data
of Risbey et al. (2009), and hence there are likely to be
intra-seasonal variations in rainfall associated with
blocking that are favoured by our analysis. Differences in
spatial resolution between gridded datasets are also
likely to contribute to these discrepancies; we use
observed rainfall data interpolated onto a 2.5 grid,
whereas Risbey et al. (2009) use data on a 0.05 grid.
Also, the NNR1 u500 data used for the blocking index in
Risbey et al. (2009) span the period 19482006, whereas
we use data for the period 19812010. Such differences
in the spatial and temporal resolution of the data are
likely to influence the depiction of significant rainfall
anomalies associated with blocking.

3.3 Rainfall anomalies


Observed composites of weekly-mean precipitation show
enhanced rainfall over most of Australia in MAM, JJA
and SON in association with blocking at 140E, with the
largest significant anomalies occurring over the southeast
of the continent (Fig. 5). Areas of reduced rainfall are
seen over northern Australia in JJA and over the northwest in SON and MAM, although these anomalies are
relatively weak. In DJF, enhanced rainfall again dominates the southeast, while a broad region of reduced
rainfall covers most of northern and western Australia
with anomalies up to -4 mm/day around the Gulf of
Carpentaria. The summer rainfall signal is clearly more
complex and perhaps influenced by the advection of
dry southern air into the tropics, which would act to
moderate the tropical monsoon flow at this time of year
(Pook et al. 2013). Further investigation also suggests
abatement of the monsoon westerlies by weak low-level

123

Simulation and prediction of blocking

Fig. 5 As for Fig. 2 but for precipitation anomalies over Australia for observations (left column), POAMA-2 (middle column), and POAMA-2
minus observed differences (right column). Contour interval is 0.5 mm/d with blue shading representing positive values

Seasonal composite maps of weekly-mean precipitation


in association with blocking at 140E are also shown in
Fig. 5 for POAMA-2 (middle column), and POAMA-2
minus observed differences (right column). We again
present composite anomalies averaged over lead times of
36 weeks since the impact of blocking on Australian
rainfall is broadly consistent from week to week in the

model. POAMA-2 reproduces well the observed blockingrainfall relationship in all seasons with spatial correlations
of 0.55 for MAM, 0.64 for JJA, 0.69 for DJF and 0.87 for
SON. The model captures the observed rainfall increase
over south-eastern Australia with comparable magnitude to
that observed in most seasons, a notable exception to this
being in JJA when models peak rainfall anomaly (1.5 mm/

123

A. G. Marshall et al.

day) is half that observed (3 mm/day). For the DJF analysis


POAMA-2 also captures the reduction in rainfall over parts
of northern and western Australia, although these anomalies are relatively weak and do not extend into the interior
of the continent as they do in the observations.
A notable difference in the POAMA-2 rainfall analysis
compared with that observed relates to the zonal expanse of
rainfall anomalies across southern Australia, particularly in
JJA and SON (incorporating the southern wet/growing
season). The observations show rainfall increases in excess
of 0.5 mm/day to extend westward from south-eastern
Australia across much of the continent; this is consistent
with a broad dynamical effect of the low z500 anomaly
over the southeast (Fig. 2). The POAMA-2 composite, on
the other hand, shows near-zero anomalies over southwestern Australia and a rainfall response that appears
rather coast-anchored in the southeast to the east of the
Great Dividing Range. This may be suggestive of orographically-derived rainfall from easterly flow transporting
moisture from the Tasman Sea, and is consistent with the
emergence of stronger easterly anomalous flow over southeastern Australia and the Tasman Sea in the model compared with that observed (Fig. 4). The primary rain-bearing
mechanism associated with blocking in POAMA-2 may
therefore differ from the observed dynamical effect of the
low z500 anomaly; a more thorough analysis is needed to
identify and understand rain-bearing mechanisms in the
model. An in-depth assessment of the synoptic systems in
POAMA-2 is planned for future research.
3.4 Roles of ENSO and SAM
ENSO and the SAM are shown to have significant impacts
on intra-seasonal rainfall variability over Australia yearround (e.g. Hudson et al. 2011a; Marshall et al. 2012a).
Therefore, the rainfall anomalies in Fig. 5 may be contaminated by the teleconnection between ENSO and Australian rainfall, and/or by the teleconnection between the
SAM and Australian rainfall, i.e. not related to blocking
alone. To test whether the blocking rainfall anomalies in
our study are influenced by either ENSO or the SAM on
intra-seasonal timescales, we first calculate the correlation
of the anomalous blocking index with indices for each
driver using weekly-mean data. For ENSO we use NNR1
weekly SST anomalies averaged over the Nino-3.4 region
of the eastern equatorial Pacific, and for the SAM we use
the Climate Prediction Center daily Antarctic Oscillation
Index1 recalculated as weekly-means. The threshold for
significance at the 95 % confidence level is calculated
using a standard t-test for the significance of correlation
1

Available at ftp://ftp.cpc.ncep.noaa.gov/cwlinks/norm.daily.aao.
index.b790101.current.ascii.

123

Table 1 Correlation of the blocking index at 140E with ENSO


(Nino-3.4 SST) and SAM indices on weekly timescales, for the period
19812010 (bold indicates significant correlation at the 95 % level;
see text for details)
ENSO

SAM

JJA

-0.13

0.03

SON

-0.12

0.29

DJF

-0.14

0.31

MAM

-0.01

0.08

Table 2 Warm and cold ENSO episodes for each season over the
period 19812010 (see text for details)
Cold ENSO

Warm ENSO

JJA

1988, 1999, 2010

1987, 1997

SON

1983, 1988, 1998, 1999,


2007, 2010

1982, 1986, 1987, 1997,


2002, 2009

DJF

19841985, 19881989,
19981999, 19992000,
20052006, 20072008,
20082009

19821983, 19861987,
19871988, 19911992,
19941995, 19971998,
20022003, 20092010

MAM

1999

1983, 1987, 1992

coefficients that replaces the number of samples N by Neff


to account for the lag 1 autocorrelation of the blocking
index at weekly timescales. Accordingly, the sample size is
effectively reduced from N = 360 to Neff = 210, and the
threshold for significance is 0.14. The high index polarity
of blocking is significantly but weakly correlated with the
cold phase of ENSO in DJF, and with the positive phase of
the SAM in SON and DJF, on weekly timescales (Table 1).
We consider the potential impact of ENSO on our
blocking-rainfall results for DJF by re-computing the
observed and model composites of Fig. 5 but excluding the
years whose seasonal ENSO index value, based on the
Climate Prediction Center Oceanic Nino Index,2 exceeds
one standard deviation about the mean for our base period
19812010 (i.e. [1r and \-1r), thereby omitting strong
warm and cold ENSO episodes from the analysis (listed in
Table 2). The result for observations is shown in the top
left panel of Fig. 6 as the difference relative to when ENSO
is present in the analysis. Predominantly, the removal of
ENSO leads to an intensification of dry anomalies over the
Top End and a southward expansion of dry anomalies into
southern Queensland, which can be anticipated considering
the influence of La Nina on rainfall in these regions (Risbey et al. 2009) in concert with the negative relationship
between ENSO and blocking (Table 1). We note, however,
that these rainfall differences do not appear statistically
2

Available at http://www.cpc.ncep.noaa.gov/products/analysis_moni
toring/ensostuff/ensoyears.shtml.

Simulation and prediction of blocking

Fig. 6 Observed weekly-mean composite differences in precipitation


over Australia for removal of ENSO in DJF (top left), the SAM in
SON (top right), the SAM in DJF (bottom left), and both ENSO and
the SAM in DJF (bottom right), relative to when both drivers are

present in the analysis (Fig. 5). Calculations are made for the highminus-low polarity of blocking at 140E over the period 19812010,
with a contour interval of 0.5 mm/d. The lack of shading indicates no
statistical significance (95 % confidence level)

significant at the 95 % confidence level. For POAMA-2,


the exclusion of ENSO years from the DJF analysis makes
no appreciable difference to the blocking-rainfall pattern
presented in Fig. 5 (not shown). We conclude that blocking
has a robust impact on Australian rainfall irrespective of
ENSO.
We also consider the potential impact of the SAM on
our blocking-rainfall results for SON and DJF by re-computing the composites of Fig. 5 but this time excluding the
years whose seasonal SAM index value, based on the
Climate Prediction Center Antarctic Oscillation Index,
exceeds one standard deviation about the mean for our base
period 19812010 (i.e. [1r and \-1r), thereby omitting
positive and negative phase SAM episodes from the analysis (listed in Table 3). The results are similar to that of
the ENSO analysis above in that the removal of the SAM
does not significantly impact the observed rainfall signal in
either season (Fig. 6), and makes no appreciable difference
to the blocking-rainfall pattern for POAMA-2 (not shown).
Again we conclude that blocking has a robust impact on
Australian rainfall irrespective of the SAM. As expected,
when we remove both ENSO and SAM years together from
the DJF analysis, the result is the reaffirmation of the

Table 3 Positive and negative phase SAM episodes for each season
over the period 19812010 (see text for details)
Negative SAM

Positive SAM

JJA

1981, 1991, 1992,


1995, 2007, 2009

1989, 1993, 1998, 2004, 2010

SON

1988, 1994, 1997,


2000, 2002

1985, 1999, 2001, 2008, 2010

DJF

19821983,
19841985,
19911992

19881989, 19931994, 19941995,


19981999, 19992000,
20012002, 20072008,
20082009

MAM

1981, 1986, 1990,


1992, 2002, 2007

1982, 1989, 1998, 1999

individual driver analyses, with no significant impact on


the rainfall signal at the 95 % confidence level (Fig. 6).

4 Mean state bias


POAMA-2 demonstrates the ability to reproduce the
anomalous large-scale structure of blocking in the Australian region (Figs. 2, 3), but underestimates the strength

123

A. G. Marshall et al.

Fig. 7 POAMA-2 minus observed difference in anomalous (top


row), climatological (second row), and total (bottom row) 500 hPa
geopotential height (left column) and 500 hPa zonal wind (right
column) in association with high index blocking in JJA over the
period 19812010. (POAMA-2 hindcasts are again averaged over

lead times of 36 weeks.) Contour interval is 20 m for geopotential


height and 1 m/s for zonal wind, with blue shading representing
negative values and white shading representing absolute values\20 m
for geopotential height and 1 m/s for zonal wind. Zero contours are
omitted from figure panels

and frequency of the total blocking index increasingly with


lead time after the first fortnight of the hindcast (Fig. 1).
We now assess the deterioration of total blocking in the
model with respect to its drifting basic state, and again
choose the JJA season for demonstration (noting that the
result is consistent for all seasons). Figure 7 presents
model-minus-observed differences for the z500 (left column) and u500 (right column) anomaly (top row), climatology (middle row), and total (bottom row) fields in
association with high index blocking. Put another way,
Fig. 7 shows model errors in the anomaly, climatology
(mean), and total fields for the z500 and u500 blocking
structure. Again we present weekly-mean composites for
the period 19812010 and average over lead times of

36 weeks, noting that the spatial patterns are again similar


at these lead times. As inferred from Fig. 2, POAMA-2
shows a weak enhancement in the blocking wave train
relative to that observed, with a 20 m increase in z500
anomalies south of Australia around 50S (i.e. a 10 %
increase in the blocking anticyclone strength, as noted in
Sect. 3b) and a decrease of up to 40 m southeast of the high
near 70S (also noted in Sect. 3b as a 25 % intensification
of the anomalous low). However by weeks 36 of the
hindcast, the models drifting basic state produces a negative z500 mean bias over much of the southern hemisphere including a local minimum (-100 m) near the
Australian blocking region, as seen in the climatology
difference panel of Fig. 7. This is accompanied by a local

123

Simulation and prediction of blocking

positive bias (40 m) southeast of the blocking region, again


centred near 70S in the vicinity of the anomalous low.
Correspondingly, a westerly mean bias appears over the
southern extremes of the continent in opposition to the
blocking easterly anomalous flow, and a strong easterly
mean bias of up to 7 m/s occurs to the southeast of the
blocking region in the vicinity of the southern branch of the
blocking split westerly flow. This mean state bias dominates the total z500 and u500 fields, acting to oppose the
weak enhancement in the anomalous blocking structure
and produce a total negative bias which is reflected in the
weakening strength and frequency of blocking at 140E
after the first fortnight (Fig. 1). This result is consistent
with Scaife et al. (2010) who also attribute large errors in
total blocking to errors in the mean state rather than errors
in the variability.
One way of addressing the issue of mean state bias in a
climate model is to use flux correction to prevent the model
from drifting into a different mean state to that observed.
We repeat the analysis of Fig. 7 for a flux-corrected version
of POAMA-2 for which climatological biases are reduced
using observed oceanatmosphere fluxes, however the
blocking bias remains (not shown). We believe that
blocking in POAMA-2 may be impacted by the models
underestimate of mean rainfall over the eastern Indian
Ocean (via the wave train teleconnection to the extra-tropics), which is a problem in both the standard and fluxcorrected versions. This relates to the convection scheme in
the model and impacts the depiction of other large-scale
intra-seasonal climate drivers such as the MJO (Marshall
et al. 2012b). Future work in this area should therefore
include model sensitivity experiments that explore the
impact of Indian Ocean rainfall biases on blocking in the
Australian region.

5 Prediction of blocking and associated rainfall


in POAMA-2
5.1 Blocking index prediction
We assess the predictive skill in forecasting the daily
blocking index at intra-seasonal timescales by first calculating the daily-mean blocking index at 140E using
absolute u500 data from observations and POAMA-2
hindcasts. We score the daily index for each hindcast
member (relative to the daily observed index) with rootmean-square error (RMSE) and correlation (COR), which
are calculated as a function of forecast start date over the
entire 30-year period, for lead times out to 30 days. RMSE
and COR calculations in Fig. 8 show that the predictive
skill of the ensemble mean hindcast (EnsMn; bold solid
line) is greater than the average skill of the individual

hindcast members (MemAv; dashed line). For a climatological forecast the RMSE = 1, and thus forecasts are
typically deemed to be skilful for RMSE \1. Figure 8 (top
row) shows the RMSE to reach a value of 1 at a lead time
of 16 days for EnsMn and 5 days for MemAv. We note
that the ensemble mean skill curve asymptotes towards a
RMSE of 1, or a COR of about 0.3.
We also wish to determine whether the model skill
outperforms the skill obtained from persistence of
observed, which is calculated by persisting the daily-mean
observed anomaly immediately prior to the forecast start
date over all lead times. Figure 8 (top row) shows that both
the skill for EnsMn and MemAv clearly outperform that of
persistence (Persist; dotted line), which has skill out to
just 2 days. Moreover, the 16-day limit of skill for the
ensemble mean hindcast far exceeds the 3-day e-folding
timescale of the persistence forecast (The e-folding timescale is determined where the persistence curve drops to a
value of COR = e-1, or COR & 0.37; 3 days is consistent
with blocking persistence timescales for the southern
hemisphere reported by Schalge et al. 2011 and OKane
et al. 2012). Thus POAMA-2 provides skilful forecasts of
the blocking index almost 2 weeks beyond the decorrelation timescale of blocking.
We are interested to know if the blocking index prediction skill varies depending on whether or not a blocking
event is occurring at the initial time, so we stratify our
observed and modelled ensemble-mean absolute blocking
index datasets into forecast start dates according to the
strength of the observed blocking index in the initial conditions. Again we define a high index as occurring when
blocking index anomalies are greater than the mean plus 1
standard deviation and a small index as occurring when
blocking index anomalies are less than half a standard
deviation about the mean, such that the high index represents physical situations for which the mid-latitude westerly flow is reduced (i.e. characteristic of strong blocking)
and the small index represents climatological mid-latitude
westerly flow. The mean and standard deviation of the
observed index, used in calculating the thresholds, are
calculated from all 1,080 hindcast start dates over the
30 years period. We find that the prediction skill extends
out to 15 days (for a RMSE of 1) irrespective of the
blocking strength at the initial time, and interestingly the
result remains unchanged when we omit ENSO and SAM
years from the analysis (Fig. 8, middle panels). This likely
suggests that Australian blocking activity does not involve
positive feedbacks, unlike for the MJO (Rashid et al. 2010)
and SAM (Marshall et al. 2012a).
We also stratify the datasets into forecast start months
DJF, MAM, JJA and SON (irrespective of blocking
strength at the initial time) to investigate how the predictive skill varies as a function of season. There is little

123

A. G. Marshall et al.

Fig. 8 Top row shows the root-mean-square error (left) and correlation (right) of the predicted blocking index at 140E for the
ensemble mean (EnsMn; bold solid line), the average of the
individual hindcast members (MemAv; dashed line), and persistence of observed (Persist; dotted line) as a function of lead time
(days) using all hindcast start dates over the period 19812010.
Second row shows the root-mean-square error (left) and correlation
(right) of the predicted blocking index at 140E for ensemble mean

hindcasts stratified according to high (solid line) and small (dotted


line) index blocking at the initial time, independent of ENSO and the
SAM over the period 19812010. Bottom row shows the root-meansquare error (left) and correlation (right) of the predicted blocking
index at 140E for ensemble mean hindcasts corrected for model drift
(solid line) compared with the original uncorrected analysis (dotted
line)

variation between seasons (not shown) with skilful prediction achieved out to about 14 days for DJF (when
blocking is weakest; Fig. 1), 15 days for MAM and SON,

and 16 days for JJA (when blocking is strongest). Finally,


we test whether the skill for predicting the blocking index
in POAMA-2 (over all start dates) is affected by the

123

Simulation and prediction of blocking

Fig. 9 Correlation skill for forecasting precipitation anomalies over


Australia during the fortnight comprising the average of weeks 2 and
3 for high (top left) and small (top right) index blocking at the initial
time, independent of ENSO and the SAM over the period 19812010.

(Statistically significant correlations are shaded; 95 % confidence


level). The difference in correlation between high and small index
blocking is also shown (bottom left; Fischers Z-transformation, 95 %
level shaded)

models mean state drift by replacing the blocking index


climatology in the model with the observed climatology to
correct for the model drift. The impact is small with the
limit of prediction skill remaining around 16 days (Fig. 8).
This may not be too surprising given that the negative drift
in the blocking index also appears relatively small up until
3 weeks lead time (Fig. 1).
Overall our analyses show that skilful prediction of the
blocking index in POAMA-2 extends to the intra-seasonal
timescale, i.e. beyond the first fortnight of the forecast.
This is a promising result considering the short timescale of
blocking and the challenges faced by numerical weather
prediction models in predicting beyond a week ahead, and
it likely reflects the ability of POAMA-2 to capture the
broad-scale circulation associated with blocking events at
these timescales (Sect. 3).

round, we now turn our interest to the ability of POAMA-2


to predict rainfall in association with blocking, for weeks 2
and 3 of the hindcast using all start dates. We omit week 1
from the analysis since deterministic forecast skill in the
first week is high due to the use of realistic atmosphere
initial conditions. Deterministic verification for the fortnight comprising the average of weeks 2 and 3 for all
forecast starts is shown in Fig. 9 in the form of correlation,
which measures the linear correspondence between the
ensemble mean forecast and observed. To assess whether
the predictive skill for rainfall is improved when the model
is initialised during episodes of strong blocking activity,
compared with when blocking is absent from the initial
conditions, we again stratify the data for high and small
blocking index anomalies at the initial time, and remove the
impacts of ENSO and SAM by again omitting El Nino, La
Nina, and positive and negative SAM years from the analysis (Tables 2, 3). Correlation skill plots for high (82 cases)
and small (203 cases) index scenarios are presented in
Fig. 9 with shading used for statistically significant correlations (95 % confidence level). The difference in correlation between high and small index blocking is also shown.

5.2 Rainfall prediction


Based on the knowledge that POAMA-2 provides skilful
prediction of the blocking index out to 16 days and a reasonable simulation of blocking rainfall over Australia year-

123

A. G. Marshall et al.

Higher skill in forecasting rainfall in weeks 2 and 3


occurs over the eastern two thirds of the continent when
blocking is high at the initial time of the hindcast. The
largest significant improvements in skill appear over
southern regions incorporating South Australia, New South
Wales and Victoria; this also holds true for each season
(not shown). We have identified blocking as having a broad
impact on Australian rainfall at intra-seasonal timescales
both in observations and in the model, with south-eastern
Australia experiencing the largest increase in weekly-mean
rainfall associated with blocking at 140E. Our results thus
demonstrate that blocking in the Australian region contributes to intra-seasonal rainfall prediction skill in POAMA-2 over much of the continent year-round, most
notably in the vicinity of the prime blocking rainfall signal.

6 Summary and concluding remarks


We assess the depiction and prediction of blocking at
140E and its impact on Australian intra-seasonal climate
variability in the Bureau of Meteorologys dynamical intraseasonal/seasonal forecast model POAMA version 2. We
focus on blocking at 140E due to its dominant influence
on growing season rainfall over the Mallee region of southeastern Australia and on synoptic systems over much of the
continent. However we note that the ability of POAMA-2
to simulate and predict blocking is consistent for blocking
at all longitudes in the Australian region. POAMA-2 simulates well the strong seasonality of the blocking index but
underestimates its strength and frequency increasingly with
lead time, particularly after the first fortnight of the hindcast, in connection with the models drifting basic state.
Our findings are consistent with other recent work in
attributing large errors in the total blocking to errors in the
mean state rather than errors in the variability, with POAMA-2 reproducing well the large-scale structure of
weekly-mean blocking anomalies in the Australian region,
including an anomalous z500 wave train structure that links
blocking at 140E with tropical variability near Indonesia.
POAMA-2 also reproduces well the observed relationship
to weekly-mean rainfall anomalies over Australia. The
separation of mean state errors from blocking frequency
errors remains challenging; while insufficient blocking in a
climate model could itself be adversely impacting the basic
state (Tibaldi and Molteni 1990; Scaife et al. 2010), ocean
model resolution can also play a role in alleviating SST
biases associated with errors in Atlantic blocking frequency (Scaife et al. 2011). As a starting point for further
investigation, we suggest that the depiction of Australian
blocking in POAMA-2 may be improved with the reduction of biases in the distribution of warm pool rainfall

123

through the tropical-extratropical wave teleconnection; this


tropical rainfall bias is a problem in both standard and fluxcorrected versions of the model.
In this study we have focussed primarily on assessing
the impacts of blocking on intra-seasonal climate anomalies in relation to Australian rainfall. In the next phase of
this research we will extend the current analysis to explore
relationships between atmospheric blocking and heat
extremes over Australia. For example, previous studies
show the establishment of a blocking high over the Tasman
Sea in association with southeast Australian heat waves
(Hudson et al. 2011c; Pezza et al. 2012). Detailed investigation into the large-scale drivers of Australian heat
waves and their representation in POAMA-2 will provide
further insight into the role of blocking on the persistence
of extreme heat over Australia.
POAMA-2 demonstrates the ability to skilfully predict
the daily blocking index at 140E out to 16 days lead time
for the ensemble mean hindcast, surpassing both the
average predictive skill of the individual hindcast members
(5 days) and the skill obtained from persistence of
observed (2 days), and exceeding the 3-day decorrelation
timescale of blocking. This ability is not dependent on the
presence of blocking in the initial conditions, since the
same 16-day lead time is found when the hindcasts are
initialised on non-blocking days. Our analyses thus show
that skilful prediction of the blocking index in POAMA-2
extends to the intra-seasonal timescale, i.e. beyond the first
fortnight of the forecast, likely reflecting the ability of the
model to capture the broad-scale circulation characteristics
of blocking at this timescale. We further show that the
prediction ability of the model is not merely due to ENSO
or the SAM, which could have been the case. This skilful
prediction of the blocking index, together with effective
simulation of blocking rainfall anomalies over Australia,
translates into higher skill in forecasting rainfall in weeks 2
and 3 over the eastern two thirds of the Australian continent when blocking is high at the initial time of the hindcast, compared to when the blocking index is small, with
the largest significant improvements in skill appearing over
South Australia, New South Wales and Victoria. Blocking
in the Australian region thus contributes to intra-seasonal
rainfall prediction skill in POAMA-2 over much of the
continent year-round, most notably in the vicinity of the
prime blocking rainfall signal.
Acknowledgments This work was supported by the Managing
Climate Variability Program of Grains Research and Development
Corporation. Thanks to Drs. Peter McIntosh, James Risbey, Terry
OKane, Kay Shelton, Meelis Zidikheri and Jorgen Frederiksen for
useful discussions throughout the course of this work. The AWAP
data were provided by the South Eastern Australian Climate Initiative
(SEACI) in Australia, and the NNR1 data were provided by the
NOAA/CIRES Climate Diagnostics Center in Boulder (USA).

Simulation and prediction of blocking

References
Alves O, Wang G, Zhong A, Smith N, Tzeitkin F, Warren G, Schiller
A, Godfrey S, Meyers G (2003) POAMA: Bureau of meteorology operational coupled model forecast system. In: Proceedings
of national drought forum, Brisbane, April 2003, pp 4956.
Available from DPI Publications, Department of Primary
Industries, GPO Box 46, Brisbane, Qld 4001, Australia
Anderson JL (1993) The climatology of blocking in a numerical
forecast model. J Clim 6:10411056
Charney JG, DeVore JG (1979) Multiple flow equilibria in the
atmosphere and blocking. J Atmos Sci 36:12051216
Colman R, Deschamps L, Naughton M, Rikus L, Sulaiman A, Puri K,
Roff G, Sun Z, Embury G (2005) BMRC Atmospheric Model
(BAM) version3.0: comparison with mean climatology. BMRC
Research Report no. 108, Bur. Met., Melbourne, Australia
Coughlan M (1983) A comparative climatology of blocking action in
the two hemispheres. Aust Met Mag 31:313
DAndrea F, Tibaldi S, Blackburn M, Boer G, Deque M, Dix MR,
Dugas B, Ferranti L, Iwasaki T, Kitoh A, Pope V, Randall D,
Roeckner E, Straus D, Stern W, Van den Dool H, Williamson D
(1998) Northern Hemisphere atmospheric blocking as simulated
by 15 atmospheric general circulation models in the period
19791988. Clim Dyn 14:385407
Gottschalck J, Wheeler M, Weickmann K, Vitart F, Savage N, Lin H,
Hendon H, Waliser D, Sperber K, Nakagawa M, Prestrelo C,
Flatau M, Higgins W (2010) A framework for assessing operational Madden-Julian Oscillation forecasts: a CLIVAR MJO
Working Group project. Bull Amer Meteorol Soc 91:12471258
Grose MR, Pook MJ, McIntosh PC, Risbey JS, Bindoff NL (2012)
The simulation of cutoff lows in a regional climate model:
reliability and future trends. Clim Dyn 39:445459
Hopkins LC, Holland GJ (1997) Australian heavy-rain days and
associated east coast cyclones: 195892. J Clim 10:621635
Hudson D, Alves O, Hendon HH, Marshall AG (2011a) Bridging the
gap between weather and seasonal forecasting: intraseasonal
forecasting for Australia. Q J R Meteorol Soc 137:673689
Hudson D, Alves O, Hendon HH, Wang G (2011b) The impact of
atmospheric initialisation on seasonal prediction of tropical
Pacific SST. Clim Dyn 36:11551171
Hudson D, Marshall AG, Alves O (2011c) Intraseasonal forecasting
of the 2009 summer and winter Australian heat waves using
POAMA. Weather Forecast 26:257279
Hudson D, Marshall AG, Yin Y, Alves O, Hendon HH (2013)
Improving intraseasonal prediction with a new ensemble generation strategy. Mon Weather Rev. doi:10.1175/MWR-D-1300059.1
Jones DA, Wang W, Fawcett R (2009) High-quality spatial climate
data-sets for Australia. Aust Meteorol Oceanogr J 58:233248
Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L,
Iredell M, Saha S, White G, Woollen J, Zhu Y, Leetmaa A,
Reynolds B, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J,
Mo KC, Ropelewski C, Wang J, Jenne R, Joseph D (1996) The
NCEP/NCAR 40-year reanalysis project. Bull Am Meteorol Soc
77:437471
Karoly DJ (1989) Southern Hemisphere circulation features associated with El Nino-Southern Oscillation events. J Clim
2:12391251
Katzfey JJ, McInnes KL (1996) GCM simulations of eastern
Australian cutoff lows. J Clim 9:23372355
Kidson JW (1988) Indices of the Southern Hemisphere zonal wind.
J Clim 1:183194
Manabe S, Holloway JL (1975) The seasonal variation of the
hydrological cycle as simulated by a global model of the
atmosphere. J Geophys Res 80:16171649

Marshall AG, Hudson D, Wheeler MC, Hendon HH, Alves O (2012a)


Simulation and prediction of the Southern Annular Mode and its
influence on Australian intra-seasonal climate in POAMA. Clim
Dyn 38:24832502
Marshall AG, Hudson D, Wheeler MC, Hendon HH, Alves O (2012b)
Evaluating key drivers of Australian intra-seasonal climate
variability in POAMA-2: a progress report. CAWCR Res Lett
7:1016
Matsueda M, Mizuta R, Kusunoki S (2009) Future change in
wintertime atmospheric blocking simulated using a 20-km-mesh
atmospheric global circulation model. J Geophys Res
114:D12114. doi:10.1029/2009JD011919
McIntosh P, Pook M, Risbey J, Hope P, Wang G, Alves O (2008)
Australias regional climate drivers. Tech Rep, Land and Water
Australia, Canberra 48 pp
Meinke H, Stone RC (2005) Seasonal and inter-annual climate
forecasting: the new tool for increasing preparedness to climate
variability and change in agricultural planning and operations.
Clim Chang 70:221253
Mo KC, White GH (1985) Teleconnections in the Southern Hemisphere. Mon Weather Rev 113:2237
Nordeng T-E (1994) Extended versions of the convective parameterization scheme at ECMWF and their impact upon the mean
climate and transient activity of the model in the tropics.
Research Dept Technical Memorandum No. 206, ECMWF,
Shinfield Park Reading RG2 9AX, UK
OKane TJ, Risbey JS, Frankze C, Horenko I, Monselesan DP (2012)
Changes in the meta-stability of the mid-latitude Southern
Hemisphere circulation and the utility of non-stationary cluster
analysis and split flow blocking indices as diagnostic tools.
J Atmos Sci. doi:10.1175/JAS-D-12-028.1
Oke PR, Schiller A, Griffin DA, Brassington GB (2005) Ensemble
data assimilation for an eddy-resolving ocean model of the
Australian region. Q J R Meteorol Soc 131:33013311
Petoukhov V, Rahmstorf S, Petri S, Schellnhuber HJ (2013)
Quasiresonant amplification of planetary waves and recent
Northern Hemisphere weather extremes. PNAS. doi:10.1073/
pnas.1222000110
Pezza AB, van Rensch P, Cai W (2012) Severe heat waves in
Southern Australia: synoptic climatology and large scale
connections. Clim Dyn 38:209224
Pook MJ (1994) Atmospheric blocking in the Australasian region in
the Southern Hemisphere winter. PhD thesis, University of
Tasmania, 168 pp
Pook MJ, Gibson T (1999) Atmospheric blocking and storm tracks
during SOP-1 of the FROST Project. Aust Meteor Mag 48:5160
Pook MJ, McIntosh PC, Meyers GA (2006) The synoptic decomposition of cool-season rainfall in the Southeastern Australian
cropping region. J Appl Meteor Climatol 45:11561170
Pook MJ, Risbey J, McIntosh P (2010) East coast lows, atmospheric
blocking and rainfall: a Tasmanian perspective. IOP Conf Ser
Earth Environ Sci 11:012011. doi:10.1088/1755-1315/11/1/012011
Pook MJ, Risbey J, McIntosh P, Ummenhofer C, Marshall AG,
Meyers GA (2013) The seasonal cycle of blocking and
associated physical mechanisms in the Australian region and
relationship with rainfall. Mon Weather Rev. doi:10.1175/
MWR-D-13-00040.1
Qi L, Leslie LM, Zhao SX (1999) Cut-off low pressure systems over
southern Australia: climatology and case study. Int J Climatol
19:16331649
Raphael MN (2004) A zonal wave 3 index for the Southern
Hemisphere. Geophys Res Lett 31:L23212. doi:10.1029/
2004GL020365
Rashid H, Hendon HH, Wheeler M, Alves O (2010) Prediction of
the Madden-Julian Oscillation with the POAMA dynamical

123

A. G. Marshall et al.
seasonal prediction system. Clim Dyn. doi:10.1007/s00382010-0754-x
Risbey JS, Pook MJ, McIntosh PC, Wheeler MC, Hendon HH (2009)
On the remote drivers of rainfall variability in Australia. Mon
Weather Rev 137:32333253
Scaife AA, Knight JR (2008) Ensemble simulations of the cold
European winter of 20052006. Q J R Meteorol Soc
134:16471659
Scaife AA, Woollings T, Knight J, Martin G, Hinton T (2010)
Atmospheric blocking and mean biases in climate models.
J Clim 23:61436152
Scaife AA, Copsey D, Gordon C, Harris C, Hinton T, Keeley S,
ONeill A, Roberts M, Williams K (2011) Improved Atlantic
winter blocking in a climate model. Geophys Res Lett
38:L23703. doi:10.1029/2011GL049573
Schalge B, Blender R, Fraedrich K (2011) Blocking detection based
on synoptic filters. Adv Meteorol ID 717812:11 pp. doi:10.1155/
2011/717812
Schiller A, Godfrey JS, McIntosh P, Meyers G (1997) A global ocean
general circulation model climate variability studies. CSIRO
Marine Research Report No 227
Schiller A, Godfrey JS, McIntosh PC, Meyers G, Smith NR, Alves O,
Wang G, Fiedler R (2002) A new version of the australian
community ocean model for seasonal climate prediction. CSIRO
Marine Research Report No. 240
Stockdale TN (1997) Coupled oceanatmosphere forecasts in the
presence of climate drift. Mon Weather Rev 125:809818
Tibaldi S, Molteni F (1990) On the operational predictability of
blocking. Tellus 42A:343365
Tiedke M (1989) A comprehensive mass flux scheme for cumulus
parameterisation in large-scale models. Mon Weather Rev
117:17791800
Toth Z, Pena M, Vintzileos A (2007) Bridging the gap between
weather and climate forecasting: research priorities for intraseasonal prediction. Bull Am Meteorol Soc 88:14271429
Trenberth KE, Mo KC (1985) Blocking in the Southern Hemisphere.
Mon Wea Rev 113:321
Uppala SM, Kallberg PW, Simmons AJ, Andrae U, Da Costa
Bechtold V, Fiorino M, Gibson JK, Haseler J, Hernandez A,
Kelly GA (2005) The ERA-40 re-analysis. Q J R Meteorol Soc
131:29613012

123

Valcke S, Terray L, Piacentini A (2000) OASIS 2.4 ocean


atmospheric sea ice soil users guide, version 2.4. CERFACS
technical report, CERFACS TR/CMGC/00-10, 85 pp
van Loon H (1956) Blocking action in the Southern Hemisphere: part
1. Notos 5:171175
van Loon H, Jenne RL (1972) The zonal harmonic standing waves in
the Southern Hemisphere. J Geophys Res 77:9921003
van Ulden AP, van Oldenborgh GJ (2006) Large-scale atmospheric
circulation biases and changes in global climate model simulations and their importance for climate change in Central Europe.
Atmos Chem Phys 6:863881
Vitart F (2004) Monthly forecasting at ECMWF. Mon Weather Rev
132:27612779
Vitart F, Buizza R, Balmaseda MA, Balsamo G, Bidlot J-R, Bonet A,
Fuentes M, Hofstadler A, Molteni F, Palmer TN (2008) The new
VarEPS-monthly forecasting system: a first step towards seamless prediction. Q J R Meteorol Soc 134:17891799
Wang G, Alves O, Smith N (2005) BAM3.0 tropical surface flux
simulation and its impact on SST drift in a coupled model.
BMRC Research Report No. 107, Bur. Met., Melbourne,
Australia
Watson JS, Colucci SJ (2002) Evaluation of ensemble predictions of
blocking in the NCEP global spectral model. Mon Wea Rev
130:30083021
Wiedenmann JM, Lupo AR, Mokhov II, Tikhonova EA (2002) The
climatology of blocking anticyclones for the northern and
southern hemispheres: block intensity as a diagnostic. J Clim
15:34593473
Woollings T, Charlton-Perez A, Ineson S, Marshall AG, Masato G
(2010) Associations between stratospheric variability and tropospheric blocking. J Geophys Res 115:D06108. doi:10.1029/
2009JD012742
Yin Y, Alves O, Oke PR (2011) An ensemble ocean data assimilation
system for seasonal prediction. Mon Weather Rev 139:786808
Zhong A, Alves O, Hendon H, Rikus L (2006) On aspects of the mean
climatology and tropical interannual variability in the BMRC
Atmospheric Model (BAM 3.0). BMRC Research Report No.
121, Bur. Met., Melbourne, Australia
Zidikheri M, Frederiksen J, OKane T (2007) Multiple equilibria and
midlatitude atmospheric blocking: a re-examination. World Sci
Lect Notes Complex Syst 6:5986

You might also like