Download Full Paper - Agronomy 2015

Impact of projected climates on drought occurrence in the Australian wheatbelt
James Watson1, Bangyou Zheng2, Scott C. Chapman2 and Karine Chenu1
The University of Queensland, Queensland Alliance for Agriculture and Food Innovation (QAAFI), 203 Tor Street, Toowoomba,
QLD 4350, [email protected]
2
CSIRO Agriculture Flagship, Queensland Bioscience Precinct, 306 Carmody Road, St. Lucia, QLD 4067
1
Abstract
Wheat is a staple crop, and in Australia it is primarily produced in rainfed environments. Climate change
projections indicate an increase in future rainfall variability and in temperature across the Australian
wheatbelt. Coupled with the continued increase in demand for this crop due to rising populations and living
standards, climate change may significantly impact the Australian wheat industry. The lead times involved in
adapting cultivars and management practices mean that planning for adaptation must often begin many years
before implementation. Thus timely, realistic assessments of the crop-level implications of climate change
are critical to the long-term planning of breeders, farmers and policy makers. Such assessments require
extensive analysis of the complex interplay between local environment, genotype, and adaptive management
practices. In this study we capture these interactions for 60 representative sites using the APSIM-Wheat
crop model, and simulated the impact that 33 climate model projections had on the distribution of drought
environment types across the Australian wheatbelt. Simulation results indicate that changes in future drought
patterns are highly region-specific. Significant variations in projected changes were found across climate
models, giving local ranges of uncertainty to consider in planning efforts. However, simulations for the
majority of climate models projected increased frequencies of severe drought conditions in the Western area
of the wheatbelt, and fewer severe droughts in other regions. Overall, simulations indicate that all areas of
the Australian wheatbelt will continue to experience drought conditions this century, and that adaptation
planning is necessary to match future wheat demand.
Key words
wheat, climate change, water deficit, environment characterisation, crop modelling, APSIM
Introduction
Due to the combination of increasing population and rising living standards, demand for staple foods
such as wheat continues to increase. Most Australian wheat is produced in water-limited environments,
and is exported. Recent climate projections indicate that Australia will experience increased temperature
and rainfall variability this century (Zheng et al. 2012; Reisinger et al. 2014), which may have substantial
implications for the local wheat industry. As it takes 5-15 years to produce new cultivars, timely assessments
of the plausible impacts of climate change are of vital importance (Chapman et al. 2012).
Both the degree and timing of environmental stressors influence crop development (e.g. Fischer 2011). Thus,
assessing the impact of abiotic stressors on wheat requires careful consideration of the complex interactions
between the local environment, crop genotype, and management practices (Chenu 2015). In this study, we
used the APSIM-Wheat model (Holzworth et al. 2014) to capture these interactions at 60 representative sites
across the Australian wheatbelt. Seasonal patterns in the water-stress index output by the model were classified
to identify key seasonal patterns of water-stress, termed ‘drought environment types’ (Chenu et al. 2013). For
each site, a set of future climate scenarios were generated using the projections of 33 climate models from the
Coupled Model Intercomparison Phase 5 Project (CMIP5; Taylor et al. 2012). The wheat model was run for
each site’s current and future climate scenarios, and every simulated season was classified into one of four
drought environment types. The main aims were to (1) assess likely changes in occurrence of drought across the
wheatbelt, and (2) identify the range of potential changes projected by this large ensemble of climate models.
Materials and Methods
Characterizing current drought environment types
Crop simulations were performed using the Agricultural Production Systems Simulator (APSIM Version 7.6)
for the wheat variety ‘Hartog’ (Triticum aestivium L.). For each of the 60 representative sites, two sets of
© 2015 “Building Productive, Diverse and Sustainable Landscapes “
Proceedings of the 17th ASA Conference, 20 – 24 September 2015, Hobart, Australia. Web site www.agronomy2015.com.au
simulations were performed using historical weather observations from 1955 to 2013, obtained from SILO
simulations
performed
historical
weather
observations
from 1955 to
2013,dates
obtained
SILO
(Jeffrey
et al. were
2001).
The first using
set was
performed
to identify
five representative
sowing
and from
five initial
(Jeffrey
et
al.
2001).
The
first
set
was
performed
to
identify
five
representative
sowing
dates
and
five
initial
soil water levels for each site, according to local conditions and management practices (see Chenu et al. 2013
soil
water levels
each dates
site, according
local
conditions
and management
practices
(see20%
Chenu
et al.
for
details).
These for
sowing
and initialtosoil
water
values were
selected to each
represent
of the
2013
for
details).
These
sowing
dates
and
initial
soil
water
values
were
selected
to
each
represent
20%
the
sowing opportunities and soil water conditions, respectively. The resulting 5x5 initial conditions for eachofsite
sowing
opportunities
and
soil
water
conditions,
respectively.
The
resulting
5x5
initial
conditions
for
each
site
were used in the second set of simulations, to characterize the drought patterns experienced by current wheat
wereacross
used inthe
thewheatbelt.
second set of simulations, to characterize the drought patterns experienced by current wheat
crops
crops across the wheatbelt.
For each model run, a daily water-stress index was computed to reflect crop ‘water supply/demand ratio’, i.e.
the
potential
water
available toindex
the crop
the amount
that the
crop
couldsupply/demand
use for potential
Forratio
eachofmodel
run,soil
a daily
water-stress
was to
computed
to reflect
crop
‘water
ratio’,
transpiration.
index ranged
fromavailable
0 (no water
available
crop) that
to 1 the
(no crop
watercould
stress).
i.e. the ratio This
of potential
soil water
to the
crop to to
thethe
amount
useWater-stress
for potential
patterns
for each
environment
(i.e.from
each0unique
site,available
year, sowing
andtoinitial
water
combination)
transpiration.
This
index ranged
(no water
to thedate
crop)
1 (no soil
water
stress).
Water-stress
were
defined
as theenvironment
water-stress(i.e.
indices
every
100°Cd,
from
aftersoil
emergence
to 450°Cd
patterns
for each
eachaveraged
unique site,
year,
sowing
date100°Cd
and initial
water combination)
after
wereflowering.
defined as the water-stress indices averaged every 100°Cd, from 100°Cd after emergence to 450°Cd
after flowering.
The ‘clara’ clustering function (Maechler et al. 2015) was used to group these water-stress patterns into four
sets to define four drought ‘environment types’ (ET; see Figure 1).
The ‘clara’ clustering function (Maechler et al. 2015) was used to group these water-stress patterns into four
sets to define four drought ‘environment types’ (ET; see Figure 1).
Figure1.1.Water-stress
Water-stresspatterns
patternsdefining
definingthe
thefour
fourmain
mainenvironment
environmenttypes
types(ET)
(ET)identified
identifiedfor
forthe
theAustralian
Australian
Figure
thth
wheatbelt.For
Foreach
eachenvironment
environmenttype,
type,the
themedian
median(coloured
(colouredline)
line)and
andvariations
variationsfrom
fromthe
the1010
the90
90thth
totothe
wheatbelt.
percentile
percentileofofthe
thedistribution
distribution(shading)
(shading)are
arepresented.
presented.
Projecting future environment types
Projecting
future
environment
typesmodels from CMIP5 (Taylor et al. 2012) were used to generate the set of
The monthly
output
of 33 climate
The
monthly
output
of
33
climate
models
from CMIP5
(Taylor
al. 2012)
weredata
usedfrom
to generate
the set of
future scenarios by downscaling daily
observations
from
SILO.etWe
employed
the Representative
future
scenarios Pathway
by downscaling
daily
observations
SILO.
We employed
data
the Representative
Concentration
(RCP) 8.5
scenario,
whichfrom
assumes
‘business
as usual’
COfrom
emissions.
For each
2
emissions.
For
each site,
Concentration
Pathway
(RCP)
8.5
scenario,
which
assumes
‘business
as
usual’
CO
site, future climate scenarios were generated according to each climate model for 2three distinct time
future
climate
scenarios
were
generated
each to
climate
threescenarios
distinct time
periods
centred
on 2030,
2050
and 2070according
(baseline:to1955
2013).model
All 99for
future
wereperiods
the result of
centred
on 2030,
2070
(baseline:
1955 to 2013).
All 99 future
scenarios
were
the result
of local
transforming
the2050
localand
daily
historical
temperature
and precipitation
values
by their
projected
future
transforming
the local
dailyclimate
historical
temperature
and precipitation
bysite,
theiryear,
projected
future
monthly means.
For each
scenario,
the water-stress
patternsvalues
of each
sowing
date,local
and
monthly
means.
each climate
scenario,
the water-stress
of each
year,
sowing date,
and
initial soil
waterFor
combination
were
simulated
as described patterns
above (with
onesite,
set of
simulations
to identify
initial
water combination
wereclimate
simulated
as described
(with one settoof
simulationsdrought
to identify
initialsoil
conditions
specific to each
model,
and one above
set of simulations
characterize
patterns).
initial conditions specific to each climate model, and one set of simulations to characterize drought patterns).
In total, 9.2 million crop simulations were performed (1 historical scenario + 99 future climate scenarios) x
sites
59 years)
+ (60
sites x 59were
years
x 5 planting
dates x 5scenario
initial soil
water
values)),
required
In((60
total,
9.2xmillion
crop
simulations
performed
(1 historical
+ 99
future
climatewhich
scenarios)
x
the sites
equivalent
of over
100sites
daysxof59computing
time on adates
current
Intel Xeon
((60
x 59 years)
+ (60
years x 5 planting
x 5 processor
initial soil(2.93GHz
water values)),
whichX5570).
requiredIn
real
time, these
were
completed in
approximately
days on(2.93GHz
the University
of Queensland’s
the
equivalent
of simulations
over 100 days
of computing
time
on a current 2.5
processor
Intel Xeon
X5570). In
high
performance
computing
facility.
Results
were
stored
as
compressed
NetCDF
files
using
the
Python
real time, these simulations were completed in approximately 2.5 days on the University of Queensland’s
1
netCDF4
packagecomputing
, and the differences
between
and theNetCDF
baselinefiles
for the
frequency
of
high
performance
facility. Results
werefuture
storedscenarios
as compressed
using
the Python
2
1
occurrence
of
different
ETs
were
processed
and
analysed
using
R
.
netCDF4 package , and the differences between future scenarios and the baseline for the frequency of
occurrence of different ETs were processed and analysed using R2.
Results and Discussion
Future projections for drought patterns varied significantly across climate models, even by 2030, primarily
1
https://github.com/Unidata/netcdf4-python
due to differences in projected precipitation. This is consistent with recent work showing that these climate
2
http://www.R-project.org/
"Building
https://github.com/Unidata/netcdf4-python
© 2015
Productive, Diverse and Sustainable Landscapes "
2
Proceedings
ofhttp://www.R-project.org/
the 17th ASA Conference, 20 – 24 September 2015, Hobart, Australia. Web site www.agronomy2015.com.au
1
© 2015 “Building Productive, Diverse and Sustainable Landscapes “
Proceedings of the 17th ASA Conference, 20 – 24 September 2015, Hobart, Australia. Web site www.agronomy2015.com.au
2
Results and Discussion
Future projections for drought patterns varied significantly across climate models, even by 2030, primarily
due to differences in projected precipitation. This is consistent with recent work showing that these climate
models
modelsbroadly
broadlyagree
agreeon
onfuture
futuretemperatures,
temperatures,but
butdo
donot
notagree
agreeon
onfuture
futureprecipitation
precipitationpatterns
patternsin
inAustralia
Australia
(Reisinger
et
al.
2014).
(Reisinger et al. 2014).
Two key aspects of these results are (1) the range of projected environment type responses, and (2)
Two key aspects
results
arethe
(1)climate
the range
of projected
environment
responses,
consensus
that canofbethese
found
among
models,
which indicate
likelytype
future
impacts.and (2) consensus
that can be found among the climate models, which indicate likely future impacts.
To
To assess
assess the
the range
range of
of impacts,
impacts, climate
climate models
models were
were ranked
ranked according
according to
to their
their projected
projected changes
changes in
in
occurrence
of
severe
drought
environments
(ET
3
and
4;
Figure
1).
Figure
2
presents
the
range
of
occurrence of severe drought environments (ET 3 and 4; Figure 1). Figure 2 presents the range of projected
projected
changes
changesin
inET
EToccurrence
occurrence for
for the
the most
most optimistic
optimistic (CESM1-BGC),
(CESM1-BGC), the
the central
central (MRI-CGCM3),
(MRI-CGCM3), and
and the
the most
most
pessimistic
pessimistic(GFDL-ESM2M)
(GFDL-ESM2M)climate
climatemodels.
models.
Figure 2. Regional averages for projected changes in the frequency of occurrence of environment types across
Figure 2. Regional averages for projected changes in the frequency of occurrence of environment types across
the wheatbelt for three future periods (rows) and three contrasting climate models (columns). The coloured
the wheatbelt for three future periods (rows) and three contrasting climate models (columns). The coloured
proportion of each box indicates the change in occurrence of each environment type compared to the baseline
proportion of each box indicates the change in occurrence of each environment type compared to the baseline
(1955 to 2013). Projections are presented for 2030, 2050 and 2070 and optimistic, central, and pessimistic climate
(1955 to 2013). Projections are presented for 2030, 2050 and 2070 and optimistic, central, and pessimistic climate
models(see
(seetext).
text).
models
Clearvariations
variationsamong
amongclimate
climatemodels
modelswere
werefound,
found,even
evenby
by2030.
2030.While
While increased
increased frequencies
frequencies of
of severe
severe
Clear
droughtconditions
conditionsin
inthe
theWest
West of
of the
the wheatbelt
wheatbelt were
were projected
projected by
by both
both the
the central
central and
and pessimistic
pessimistic climate
climate
drought
models,the
theoptimistic
optimisticclimate
climatemodel
modelprojected
projectedaadecrease
decreasein
inthese
theseconditions.
conditions.In
Incontrast,
contrast,in
inthe
theEast
Eastand
and
models,
South-Eastof
ofthe
thewheatbelt,
wheatbelt,the
theoptimistic
optimisticand
andcentral
centralclimate
climatemodels
modelsindicated
indicatedaadecrease
decreasein
insevere
severedrought
drought
South-East
conditionslater
laterthis
thiscentury.
century. This
This reduction
reduction was
was primarily
primarily driven
driven by
by shorter
shorter crop
crop cycles
cycles arising
arising from
from higher
higher
conditions
temperatures,and
andincreased
increasedtranspiration
transpirationefficiency
efficiency due
due to
to raised
raised CO
CO22 concentrations.
concentrations. Note
Note that
that the
the primary
primary
temperatures,
differences in
in the
the future
future drought
drought predictions
predictions presented
presented in
in Figure
Figure 22 derive
derivemore
morefrom
fromthe
theclimate
climatemodel
model
differences
consideredthan
thanfrom
fromthe
theprojection
projectiontime.
time.
considered
When assessing the projections of all 33 climate models, occurrences of severe drought environments (ET 3
When assessing the projections of all 33 climate models, occurrences of severe drought environments (ET
and 4) were found to decrease this century in the East, South and South-East of the wheatbelt (Figure 3). In
3 and 4) were found to decrease this century in the East, South and South-East of the wheatbelt (Figure 3).
contrast, the majority of climate models projected that the Western area of the wheatbelt will experience
In contrast, the majority of climate models projected that the Western area of the wheatbelt will experience
significant
significant increases
increases in
in drought
drought conditions.
conditions.
© 2015 "Building Productive, Diverse and Sustainable Landscapes "
Proceedings of the 17th ASA Conference, 20 – 24 September 2015, Hobart, Australia. Web site www.agronomy2015.com.au
© 2015 “Building Productive, Diverse and Sustainable Landscapes “
Proceedings of the 17th ASA Conference, 20 – 24 September 2015, Hobart, Australia. Web site www.agronomy2015.com.au
3
Figure 3. Projected occurrence of drought environments (ET 3 and 4) per major wheat-producing area,
Figure
3. Projected
occurrence
of drought
(ETsummarize
3 and 4) per
wheat-producing
area,for each of
compared
to historical
occurrence
(dottedenvironments
lines). Boxplots
themajor
distribution
of percentages
compared
to
historical
occurrence
(dotted
lines).
Boxplots
summarize
the
distribution
of
percentages
for each of
the 33 climate models.
the 33 climate models.
Conclusions
Climate model projections indicate that the West of the Australian wheatbelt is likely to experience increased
Conclusions
droughtmodel
conditions
as early
as 2030.
the lead
times
involved
for crop is
adaptation,
and the western
Climate
projections
indicate
thatGiven
the West
of the
Australian
wheatbelt
likely to experience
increased
wheatbelt’s
significant
contribution
towards
total
national
production,
these
results
provide
a strong
case
drought conditions as early as 2030. Given the lead times involved for crop adaptation, and the
western
for immediate
initiation
of adaptation
planning
this area.
Other areas
ofresults
the wheatbelt
projected
wheatbelt’s
significant
contribution
towards
totalfor
national
production,
these
provideare
a strong
caseto
for
experience
decreased
frequencies
of
drought
conditions,
with
rising
temperatures
shortening
the
crop
immediate initiation of adaptation planning for this area. Other areas of the wheatbelt are projected to cycle
and CO2 concentrations
allowingofmore
efficient
water use.
even whereshortening
the frequencies
of severe
experience
decreased frequencies
drought
conditions,
withHowever,
rising temperatures
the crop
cycle
drought
conditions
are
projected
to
decrease,
their
occurrences
are
projected
to
remain
substantial.
Overall,
and CO2 concentrations allowing more efficient water use. However, even where the frequencies of severe
increased
demandare
forprojected
wheat means
that Australia
must continue
improvetocultivars
management
drought
conditions
to decrease,
their occurrences
aretoprojected
remain and
substantial.
Overall,
practises
to
mitigate
the
drought
conditions
experienced
across
its
wheatbelt.
increased demand for wheat means that Australia must continue to improve cultivars and management
practises to mitigate the drought conditions experienced across its wheatbelt.
References
Chapman, S.C., et al. 2012. Plant adaptation to climate change – opportunities and priorities in breeding.
References
Crop and Pasture Science. 63:251-268.
Chapman,
S.C.,
et 2013.
al. 2012.
Plant adaptation
to climate
opportunities
and priorities
in breeding.
Chenu, K.,
et al.
Large-scale
characterization
ofchange
drought– pattern:
a continent-wide
modelling
Crop
and
Pasture
Science.
63:251-268.
approachapplied to the Australian wheatbelt – spatial and temporal trends. New Phytologist 198:801-820.
Chenu,
al. 2013.
Large-scalethe
characterization
of drought
pattern:
a continent-wide
modelling
approach
Chenu,K.,
K.et2015.
Characterising
crop environment
– nature,
significance
and applications.
In: SadrasV.O.,
applied
to the
Australian
– spatialApplications
and temporalfor
trends.
New
Phytologist and
198:801-820.
Calderini,
D.F.,
editors. wheatbelt
Crop Physiology.
Genetic
Improvement
Agronomy.
Chenu,
K.
2015.
Characterising
the
crop
environment
–
nature,
significance
and
applications.
In: Sadras
Academic Press. pp. 321-348.
V.O., Calderini,
editors.
Crop Physiology.
for Genetic
Improvement
and
Fischer,
R.A. 2011.D.F.,
Wheat
physiology:
a review ofApplications
recent developments.
Crop
& Pasture Science
62:95–114.
Agronomy.D.P.,
Academic
Press.
pp. 321-348.
Holzworth,
et al. 2014.
APSIM
- Evolution towards a new generation of agricultural systems
Fischer,
R.A. 2011.
Wheat physiology:
a review
of recent
developments. Crop & Pasture Science 62:95–114.
simulation.
Environmental
Modelling
& Software.
62:327-350.
Holzworth,
D.P.,
et 2001.
al. 2014.
APSIM
- Evolution
towards
a new generation
of agricultural
Jeffrey, S.J.,
et al.
Using
spatial
interpolation
to construct
a comprehensive
archive systems
of
simulation.
Environmental
Modelling
&
Software.
62:327-350.
Australianclimate data. Environmental Modelling and Software. 16:309-330.
Jeffrey,
S.J.,A.
et et
al.al.
2001.
Using
spatial interpolation
to construct
comprehensive
archive
Australian
Reisinger,
2014:
Australasia.
In: Climate Change
2014:a Impacts,
Adaptation,
andofVulnerability.
climate
Environmental
Modelling and
Software.Group
16:309-330.
PartB:data.
Regional
Aspects. Contribution
of Working
II to the Fifth Assessment Report of the
Reisinger,
A. et al. 2014:
Australasia.
In: Climate
2014:
Impacts, Press,
Adaptation,
and Vulnerability.
Intergovernmental
Panel
on Climate
Change. Change
Cambridge
University
Cambridge,
UK and NewPart
B:York,
Regional
Aspects.
NY, USA,
pp. Contribution
1371-1438. of Working Group II to the Fifth Assessment Report of the
Intergovernmental
Panel on
Climate
Change.
Cambridge
Press,RCambridge,
UK and
New
Maechler,
M., et al. 2015.
Cluster:
Cluster
Analysis
BasicsUniversity
and Extensions.
package version
2.0.1.
York, K.E.,
NY, USA,
pp. 1371-1438.
Taylor,
et al. 2012:
An Overview of CMIP5 and the experiment design. Bulletin of the American
Meteorological
Society.
93:485-498.
Maechler,
M., et al. 2015.
Cluster:
Cluster Analysis Basics and Extensions. R package version 2.0.1.
Taylor,
K.E.,
et
al.
2012:
An
Overview
CMIP5what
and are
the the
experiment
offrost
the American
Zheng, B., et al. 2012. Breeding for theoffuture:
potentialdesign.
impactsBulletin
of future
and heat
eventson sowing
and flowering
time requirements for Australian bread wheat (Triticum aestivium)
Meteorological
Society.
93:485-498.
varieties?
Biology.
Zheng,
B., et al.Global
2012. Change
Breeding
for the 18:2899-2914.
future: what are the potential impacts of future frost and heat events
on sowing and flowering time requirements for Australian bread wheat (Triticum aestivium) varieties?
Global Change Biology. 18:2899-2914.
© 2015 "Building Productive, Diverse and Sustainable Landscapes "
Proceedings of the 17th ASA Conference, 20 – 24 September 2015, Hobart, Australia. Web site www.agronomy2015.com.au
© 2015 “Building Productive, Diverse and Sustainable Landscapes “
Proceedings of the 17th ASA Conference, 20 – 24 September 2015, Hobart, Australia. Web site www.agronomy2015.com.au
4