holger meinke and colleagues dpi - apsrujhansen/meinke25july.pdf · holger meinke and colleagues...
Post on 21-Apr-2020
10 Views
Preview:
TRANSCRIPT
Holger Meinke and colleagues
DPI - APSRU
Holger Meinke and colleagues
DPI - APSRU
From Decision to Discussion Support –Decision Making Under Uncertainty
From Decision to Discussion Support –Decision Making Under Uncertainty
Data < Information < Knowledge < Wisdom Data < Information < Knowledge < Wisdom
A perfect forecast is also perfectly useless unless it changes a
decision
A perfect forecast is also perfectly useless unless it changes a
decision
Systems Research at APSRUOverview
Systems Research at APSRUOverview
ρ Setting the sceneρ Cropping system context - NE Australiaρ Background on APSRU - 1990-2000ρ APSRU’s research approach and outcomes:
• systems research philosophy• key result areas for 2001-2005; research issue examples• core technologies
The ‘triple bottom line’The ‘triple bottom line’
Using the best science has to offer we need to improve the performance of agricultural systems in terms of
ρ economic performanceρ environmental impactρ and social consequences
effective integration (of data, information or knowledge) requires an issue focus rather than a methodology or
discipline focus
Current practices:low profitability, highenvironmental impact
Current practices:Current practices:low profitability, highlow profitability, highenvironmental impactenvironmental impact
Future practices:high profitability, low environmental impactFuture practices:Future practices:high profitability, low environmental impacthigh profitability, low environmental impact
Technologies must match the socio-economic system ==> need
for rural sociology
Technologies must match the socio-economic system ==> need
for rural sociology
Threshing finger millet, India
Threshing wheat, Australia
ContextThe NE Australian cropping system
Characterised by• high climate variability• cost-price squeeze• erodible soils• declining fertility• policy of self-reliance• a wide range of crop options• water main determinant of crop yield • soil water storage by fallowing• fixed rotations vs opportunity cropping • some ability to forecast seasons • concerns over salt and pesticides
APSRU Background 1990-2000APSRU Background 1990-2000Established 1990 to capture synergies between DPI and CSIRO in development and application of agricultural models• initially 15 staff co-located in Toowoomba - regional focus• joint venture arrangement with 5-year horizon• external review and renewal process 1995 and 2000• now 70 staff from 4 organisations (DPI, NR&M, CSIRO, UQ)• national focus with international linkages • development of cropping system modelling platform (APSIM) as core
technology
Practice -relevance-
Research-understanding-
Modelling-integration-
APSRU’s Systems Research Philosophy- Integrated Systems Approach -
APSRU’s Systems Research Philosophy- Integrated Systems Approach -
To benefit subtropical Australia, and the nation, through innovative agricultural systems R&D in rural industries
What is a systems approach?What is a systems approach?
ρ network of interacting elements receiving certain inputs and producing certain outputs where the systems dynamics are implicit
ρ can be applied to the full range of agricultural and natural ecosystems and their associated business and government systems, climate or economic systems.
An effective systems approach...An effective systems approach...ρ clearly defines the boundaries of the systemρ explicitly addresses the scale of the system
(what is internal / external to the system?) andρ seeks and utilises understanding of system
composition and dynamics derived from relevant research to enable prediction of the responses or behaviour.
...will include all relevant agricultural and natural ecosystems components and their associated business and government structures, eg.
• a crop or field and its management• a farm and its management• the national wheat crop and its marketing• the national drought policy and its development• a catchment and its management• a species population and its management
For any well defined issue, an effective systems approach...For any well defined issue, an effective systems approach...
APSRU’s Key Result Areas (KRA)APSRU’s Key Result Areas (KRA)KRA 1: Sustainable Agricultural Systems: Agricultural sustainability positively
enhanced via innovative research in rural business systems
KRA 2: Sustainable Landscapes: New approaches towards integrating and balancing environmental, social and economic factors that can contribute to more sustainable management of landscapes.
KRA 3: Improved Crop Design For Productivity And Sustainability:Enhanced efficiency and effectiveness in plant breeding and cropimprovement programs
KRA 4: Modelling Tools and Methods: Research communities and industry benefiting from and utilizing a range of quality software tools
KRA 5: APSRU Communication And Collaboration: APSRU functioning as a high achieving research team through coordination of its activities and broad collaboration with its stakeholders.
ρ Issue - how to manage risks generated by climate variability
ρ Research questions - can agricultural systems be managed better by improved
analysis of risks?- can decision makers value simulation as a tool in this
process?- can seasonal climate forecasts be used to advantage?
Risk ManagementRisk ManagementKRA I - Sustainable Agricultural SystemsKRA I - Sustainable Agricultural Systems
Several conditions must be met before a seasonal forecast can result in improved
outcomes. A forecast must be
Several conditions must be met before a seasonal forecast can result in improved
outcomes. A forecast must be
ρ skilful, ρ probabilistic (ie. ‘honest’), ρ relevant (neither trivial nor obvious; timely),ρ of value, ρ and the information content must be applied.
‘Skill’ vs ‘Value’‘Skill’ vs ‘Value’
ρ Clearly define what we mean by ‘skill’ρ Move towards a consensus view why we need to
measure ‘skill’ and how to do it ρ Put ‘skill’ into perspective together with other
essential attributes of effective forecasting systems
Some factsSome facts
ρ ‘Skill’ does not linearly translate from one forecast quantity to the next (eg. rainfall not equal production)
ρ ‘Skill’ is an essential but not a sufficient attribute of a effective forecasting system
ρ ‘Skill’ must be seen in context with the decisions based on the forecast
SHIFT
DISPERSION
Reference distributions
Forecast distribution
Reference distributions
Forecast distribution
KL1
ICCP
SAVGD
KWP
AMD
LEPSP
PEX
QKL
LRP
VR
LEPSBS
KL2RRIQR
BSP
Quality Measures
Shift (3)Dispersion (4)
Traditional measures (4)Other (7)
Measure of shift in QKL
Intra Class Correlation
Squared Average Deviation
Kruskall-Wallis test p- value
Absolute Median Difference
Revised LEPS score (adapted)
Probability of exceedence
Kullback-Leibler Divergence measure (KL1 + KL2)
Log rank test p-value
Variance Ratio
Revised LEPS scoreBrier score
Measure of variance in QKL80% Range RatioInter Quartile Range Ratio
Brier score (adapted)
***All quality measures were derived using cross validation
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Shift
Dis
pers
ion
Q1
Q2Q3
Q4
Comparison of Forecast Quality Measures –Wheat
Comparison of Forecast Quality Measures Comparison of Forecast Quality Measures ––WheatWheat
No single quality measure is adequate!!!
Symbols: �- shift, - dispersion, - traditional measures and - QKL
To effectively manage risk in a variable climate requires
To effectively manage risk in a variable climate requires
ρ a sound scientific understanding of the causes of climatic variability, the consequences for production and our ability to predict future variability,
ρ knowledge and appreciation of how this ability to forecast production risks can influence and change management decisions and
ρ ability to implement this knowledge to achieve improved outcomes.
Name and/or Type of Climate Phenomena Frequency (years)
Madden-Julian Oscillation, intraseasonal (MJO) 0.1 – 0.2 SOI phases based on El Niño – Southern Oscillation (ENSO), seasonal to interannual
0.5 – 7
Quasi- biennial Oscillation (eg. NAO) 1 – 2 Antarctic Circumpolar Wave (AWC), interannual 3 – 5 Latitude of Sub-tropical ridge, interannual to decadal ?? – 11 Interdecadal Pacific Oscillation (IPO) 13+ Decadal Pacific Oscillation (DPO) 13 – 18 Multidecadal Rainfall Variability 18 – 39 Interhemispheric Thermal Contrast (secular climate signal) 50 – 80 Climate change ???
Agricultural Systems and Climate Variability
Agricultural Systems and Climate Variability
Decision Type (eg. only)Logistics (eg. scheduling of planting / harvest operations)Tactical crop management (eg. fertiliser / pesticide use)Crop type (eg. wheat or chickpeas)Crop sequence (eg. long or short fallows)Crop rotations (eg. winter or summer crops)Crop industry (eg. grain or cotton, phase farming)Agricultural industry (eg. crops or pastures)Landuse (eg. agriculture or natural systems)Landuse and adaptation of current systems
Frequency (years)Intraseasonal (> 0.2)Intraseasonal (0.2 – 0.5) Seasonal (0.5 – 1.0)Interannual (0.5 – 2.0)Annual / biennial (1 – 2) Decadal (~ 10)Interdecadal (10 – 20)Multidecadal (20 +)Climate change
Dalby Rainfall
0
100
200
300
400
50018
8818
9719
0019
0519
1219
1419
4019
4119
4619
5119
6919
7719
8719
9119
93
Consistently negative April/May SOI Phase
June
to O
ct R
ainf
all (
mm
)
Median June to Oct Rainfall (190)
Dalby Wheat Yields
0123456789
1888
1897
1900
1905
1912
1914
1940
1941
1946
1951
1969
1977
1987
1991
1993
Yiel
d (t/
ha)
Median Wheat Yield
Quantifying climate variability: the importance of agricultural simulation models
Models are needed to objectively assess the impact of climate variability, seasonal climate
forecasting and management on agricultural systems
Dalby Rainfall
0
100
200
300
400
500
1888
1897
1900
1905
1912
1914
1940
1941
1946
1951
1969
1977
1987
1991
1993
Consistently negative April/May SOI Phase
June
to O
ct R
ainf
all (
mm
)
Median June to Oct Rainfall (190)
Dalby Wheat Yields
0123456789
1888
1897
1900
1905
1912
1914
1940
1941
1946
1951
1969
1977
1987
1991
1993
Yiel
d (t/
ha)
Median Wheat Yield
Dalby Rainfall
0
100
200
300
400
500
1893
1909
1916
1930
1942
1952
1962
1965
1970
1978
1980
1983
1990
Rapidly rising April/May SOI Phase
June
to O
ct R
ainf
all (m
m) Median June to Oct Rainfall (190 mm)
Dalby Wheat Yields
0123456789
1893
1909
1916
1930
1942
1952
1962
1965
1970
1978
1980
1983
1990
Yiel
d (t/
ha)
Median Wheat Yield
‘Skill’ does not linearly translate from one forecast quantity to the next
Better on-farmrisk management
Cropping SystemsModelling
Seasonal ClimateForecasting
APSI
M
M anager
BiologicalM odules
S urface Residue
E nvironmentalM odules
E rosionB
E rosionA
O ther N moduleorS oilN
CropC
CropB
CropA
P astureC
P astureB
P astureA
S wimorS oilwat
E conomics Climate
Workshops
• Analysis of risky crop management decision options• Designed with public and private advisors to suit their needs
An APSIM derived tool:
Whopper Cropper- Crop Management Discussion Support -
Whopper Cropper- Crop Management Discussion Support -
� Generates analyses for input to discussion on management issues� Data base of APSIM simulations with simple presentation graphics
� Analyses of range of crop management issues� Analyses of range of soil conditions � Analyses of utility of seasonal climate forecasts
� Which crop to sow?� When to sow?� How much N to apply?� Which variety to sow?� What density?� Analysis of different starting
conditions and seasonal forecasts
Exploring "What if?" questions:
Whopper Cropper -big decisions made easy
Graphical outputs
Applied NO3
0 kg per ha 25 kg per ha 50 kg per ha 75 kg per ha 100 kg per ha 150 kg per ha
Yiel
d (k
g/ha
)
5000
4500
4000
3500
3000
2500
2000
1500
1000
Applied NO3=50 kg per ha;SOI Phase=NegativeApplied NO3=50 kg per ha;SOI Phase=Positive
GM ($/ha)
350300250200150100500-50-100-150
Prob
abili
ty
90
80
70
60
50
40
30
20
10
3010.924NameApplied NO3=50 kg per haYield (kg/ha) Over Time
Year
19901980197019601950194019301920191019001890
Yield
(kg/
ha)
5000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
Sorghum yield – time seriesSorghum yield – time series3010.924NameApplied NO3=50 kg per haYield (kg/ha) Over Time
Year
19901980197019601950194019301920191019001890
Yie
ld (k
g/ha
)
5000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
Maturity * density * SOI (Emerald)
Moderate depth vertisol, full profile, Nov planting
Managing Water & N in a Variable Climate
Box plot of Yield (kg per ha)
Variety
Late
Early
Yiel
d (k
g pe
r ha)
7000
6000
5000
4000
3000
2000
1000
0
Late maturityHigh density
Early maturityLow density
SOI phase consistently negative SOI phase consistently positiveBox plot of Yield (kg per ha)
SOI Phase,Variety
Late
Nega
tive
Early
Nega
tive
Yiel
d (k
g pe
r ha)
7000
6000
5000
4000
3000
2000
1000
0
Late maturityHigh density
Early maturityLow density
Box p lot of Yi eld (kg pe r ha)
SOI Phase,V ariety
Late
Pos
itive
Ear
lyP
ositi
ve
Yie
ld (k
g pe
r ha)
7000
6000
5000
4000
3000
2000
1000
0
Late maturityHigh density
Early maturityLow density
ρ Basis of simulation-aided discussion - process not toolρ Private and public advisors being trained
Maturity * density * SOI (Emerald)
Moderate depth vertisol, full profile, Nov planting
Managing Water & N in a Variable Climate
#
#
#
#W A
NT
SA
NSW
VIC
TAS
R om a
D al by
Em e r al d
G o ond i wi n di
Legend:0-10%10-20%20-30%30-40%40-50%50-60%60-70%70-80%80-90%90-100%N o data
Probability of exceeding median shire wheat yieldSOI phase “consistently negative”, May/June
Commodity forecasting
What type of model do we need?What type of model do we need?• Depends on the issue and the clients (customer
focus)• Needs to deliver agreed outcomes (achievable)• Needs to capture consequences of key drivers
(relevant)
‘21st Century Transport’ might not be the best example….
Modelling as a communication toolModelling as a communication toolρThe language of climatologistsDecadal climate variability is evident in the first EOF of near-global SSTs and Pressure Field data, band filtered for various frequencies.
Decadal Variability - EOF scores
-3
-2
-1
0
1
2
3
1860 1885 1910 1935 1960 1985
11-13 years 15-20 years
Modelling as a communication toolModelling as a communication toolρThe language of farmersThe last few years were bad - first the drought and then too much rain. What can I do to better cope with such a variable climate?
25
Decadal Variability - EOF scores
-3
-2
-1
0
1
2
3
1860 1885 1910 1935 1960 1985
11-13 years 15-20 years
Modelling as a communication toolModelling as a communication toolρThe language of crop physiologistsCrop transpiration demand (Td) is determined by the amount of intercepted radiation and VPD
Td = TEc / VPD * (RUE * I)(TEc_Wheat = 4.7 g m-2 mm-1 kPa)
0
400
800
1200
1600
0 200 400 600 800
Cumulative intercepted PAR (MJ m-2)
TDM
(g m
-2)
Decadal Variability - EOF scores
-3
-2
-1
0
1
2
3
1860 1885 1910 1935 1960 1985
11-13 years 15-20 years
Modelling as a communication toolModelling as a communication toolρ The language of computer programmers! potential biomass production based on intercepted radiation
call dm_plt_tot_pot () ! demand for soil water. new subroutine sw_demand uses ! dm_plt_tot_act_dlt to determine demand for soil water based on ! a tec of 4.7
call sw_demand (c_demand_switch)
Decadal Variability - EOF scores
-3
-2
-1
0
1
2
3
1860 1885 1910 1935 1960 1985
11-13 years 15-20 years
Modelling as a communication toolModelling as a communication toolρThe language of cropping systems scientistsChanging between dryland cotton and intensive cereal rotations based on the long-term climate outlook could lift your average profits by 50%. This would also reduce your runoff and erosion potential.
Decadal Variability - EOF scores
-3
-2
-1
0
1
2
3
1860 1885 1910 1935 1960 1985
11-13 years 15-20 years
Median wheat yields and standard deviations by April/May SOI phase
0
1
2
3
4
5
Con neg Con pos Rap fall Rap rise Near '0'
Yiel
d (t
/ ha
)(eg: 1997)
(eg: 1998)
SOI Phases
Std Dev
MeanMedian
Modelling as a communication toolModelling as a communication tool
Decadal Variability - EOF scores
-3
-2
-1
0
1
2
3
1860 1885 1910 1935 1960 1985
11-13 years 15-20 years
Medi an wheat yi elds and stand ard d eviation s by April /M ay SOI ph ase
0
1
2
3
4
5
Con neg Con pos Rap fall Rap r ise Near '0 '
Yiel
d (t
/ ha)
(eg: 1997)
(eg: 1998)
SOI Phases
Std DevMeanMedian
The aim is to provide knowledge and wisdom systems rather than information systems
Vision without action is just talk.Vision without action is just talk.
Action without vision is just activity.
Vision combined with action can change the world.
KRA I - Sustainable Agricultural SystemsKRA I - Sustainable Agricultural Systemsρ System design for sustainability
- paddocks over long term - soil health; solute and pesticide movement
ρ Risk management- paddocks over short term- production tactics; climate risk
ρ Whole farm management- business unit focus- enterprise mix; spatial aspects
System design for sustainabilitySystem design for sustainability
ρ Issue - dryland salinity from rising water tableρ Research question
- can drainage and solute movement be reduced and profitability retained by re-design of the cropping system?
KRA I - Sustainable Agricultural SystemsKRA I - Sustainable Agricultural Systems
Problem
Simulation analysis
Introduction of Lucerne?
Field experimentation
Research on dryland salinityResearch on dryland salinity
Change cropping intensity?
0
50
100
150
200
250
300
350
400
450
01-Apr-93
30-Sep-93
01-Apr-94
30-Sep-94
01-Apr-95
30-Sep-95
31-Mar-96
29-Sep-96
31-Mar-97
29-Sep-97
31-Mar-98
29-Sep-98
Date
Rai
n / E
vapo
ratio
n (m
m)
RainMeasured ETPredicted ET
Evapotranspiration: wheat / lucerneEvapotranspiration: wheat / lucerne
93 Wheat
Fallow
94 Lucerne
Fallow
95-97 Lucerne
Drainage below the root zonefor four simulated crop rotations on the eastern Darling Downs
0
1000
2000
3000
Years
Wheat zero till
Sunflower zero till
Wheat / Sunflower
Wheat / bare fallow
20 40 60
Drainage (mm)
0
KRA II - Sustainable LandscapesKRA II - Sustainable Landscapesρ Vegetation and water management
- regional salinity and water resource management- links to landscape modelling
ρ Regional scale operations- industry and supply chain links- commodity forecasting; finance; insurance
Pesticide movement studiesPesticide movement studies
0
200
400
600
800
1-Nov-94 1-Dec-94 31-Dec-94 30-Jan-95 1-Mar-95 31-Mar-95Date
Soil concentration
(g/ha)
MeasuredSimulatedSprays
High risk periods
Endosulfan in cotton fields
0
5
10
15
20
25
Conventional Improvedirrigation
Dryland Stubbleretained
Reducedsprays
Annual endosulfan transport
(g/ha)
Site: Emerald
0.3%of applied
0.1%
0.03%0.1%
0.4%
Pesticide movement studiesPesticide movement studiesMeasured movement according to management practice
Land management in a catchment contextLand management in a catchment context
ρ Issue - connecting studies on salt and pesticide movement at field/farm level to catchment level
KRA III - Organism DesignKRA III - Organism Designρ Matching organisms to environment
- Genotype x Management x Environment optimisation (G x M x E)
- physiological dissection and modelling of trait variation- links to plant breeding and gene technologies
Environment Characterisation Studies- Crop View of Abiotic Stress -
Environment Characterisation Studies- Crop View of Abiotic Stress -
Relative transpiration patterns through the season
0
0.2
0.4
0.6
0.8
1
0 400 800 1200
Thermal time from emergence (°Cd)
Stress index(water supply/
demand)
Mild terminal stressSevere terminal stressMid-season stress
Clustered stress indices to produce three environment types from 547 simulations
Environment Characterisation Studies- Crop View of Abiotic Stress -
Environment Characterisation Studies- Crop View of Abiotic Stress -
Relative frequency of environment types
0%
25%
50%
75%
100%
1891 1903 1915 1927 1939 1951 1963 1975 1987Start year of 12 year sequence
Frequency of Drought
Environment Type
3.0
3.5
4.0
4.5
5.0
Yield(t/ha)
Mild terminal Severe terminal Mid-season Yield
Overall ratios of37:35:28
M-T:S-T:M-S
Changes in frequencies effect yield likelihood and cause G x EImportant in weighting representativeness of selection environments
Trait Simulation StudiesTrait Simulation Studies
ρ Estimate value of trait in particular environmentsρ Examine breeding strategies by connecting to models
of breeding systems to simulate trajectories on adaptation landscape
KRA IV - Modelling Tools & ApproachesKRA IV - Modelling Tools & Approachesρ Software engineering
• APSIM• derived products eg Whopper Cropper, DamEasy• methodology development - landscape linkages; gene
function linkages; optimisation
Agricultural Production Systems Simulator (APSIM)
Simulatesρ yield of crops and pastures ρ key soil processes (water, N, carbon)ρ surface residue dynamics & erosionρ range of management options ρ crop rotations + fallowingρ short or long term effectsρ BUT, not pests nor diseases
APSIM aims to deliver ….ρ A solid platform for modelling system’s performance over
timeρ Equal emphasis on crop and soil dimensions ρ Capability to deal comprehensively with management
mattersρ A new emphasis on better software engineering process to
achieve the above
Key features of agricultural systems
Runoff, erosion,chemicals
Crop production
Soil and residue management
Drainage
Evaporation
Soil water
Transpiration
T I M E
Soil nutrients
Crop managementWeeds
What type of model do we need?What type of model do we need?• Depends on the issue and the clients (customer
focus)• Needs to deliver agreed outcomes (achievable)• Needs to capture consequences of key drivers
(relevant)
‘21st Century Transport’ might not be the best example….
Dimensions of Systems Models
ParticipatoryAction
ResearchApplications with / for clients
ScienceBio-physical
ModulesEngineering
Module interfaces,User interfaces, Utilities
Databases etc
Researchvia
experimentation
ImprovedSoftware
EngineeringProcess
Education and teaching
Successful systems analytical approaches are based on
Successful systems analytical approaches are based on
ρ client involvement relevanceρ good scienceρ sound and efficient implementation (software
engineering) Manager
Report
Soil pH
Soilwater
Soil N
Erosion
Surface Residue
ENGINE
Crop CMaizWheat
Crop CCrop BCowpea
Soil P
Arbitrator
Climate
top related