AgMIP aims to improve agricultural modeling and assessment of climate change impacts and adaptation. It brings together regional experts to build capacity for climate, crop, and economic modeling. Key objectives include developing consistent protocols for regional climate impact simulations and assessments, improving regional models, and identifying and prioritizing adaptation strategies. AgMIP takes a two-track science approach, improving models through intercomparison and conducting multi-model climate change assessments. It focuses on major agricultural regions globally.
2. Why AgMIP?
• Agricultural risks growing, including climate
change
• Consistent approach needed to enable
agricultural sector analysis across relevant
scales and disciplines
• Long-term process lacking for rigorous
agricultural model testing, improvement, and
assessment
2
3. AgMIP Objectives
• Improve scientific and adaptive capacity of major agricultural regions
in developing and developed world
• Collaborate with regional experts in agronomy, economics, and
climate to build strong basis for applied simulations addressing key
regional questions
• Develop framework to identify and prioritize regional adaptation
strategies
• Incorporate crop and agricultural trade model improvements in
coordinated regional and global assessments of future climate
conditions
• Include multiple models, scenarios, locations, crops and participants
to explore uncertainty and the impact of methodological choices
• Link to key on-going efforts
– CCAFS, Global Futures, MOSAICC, National Adaptation Plans
3
4. AgMIP Two-Track Science Approach
Data at
Sentinel Sites
Platinum
Gold
Silver
Track 1: Model Improvement and Intercomparison
Track 2: Climate Change Multi-Model Assessment
Cross-Cutting Themes: Uncertainty, Aggregation Across Scales*, Representative Agricultural
Pathways
5. AgMIP Regions
45˚
0˚
-45˚
-90˚ 0˚ 90˚
Benefits include:
- Improved capacity for climate, crop, and economic modeling to
identify and prioritize adaptation strategies
- Consistent protocols and scenarios
- Improved regional assessments of climate impacts
- Facilitated transdisciplinary collaboration and active partnerships
- Contributions to National Adaptation Plans
6. Crop Model Pilot Activities in
AgMIIP
Crop Modeling Coordinators
K. J. Boote, Univ. of Florida
Peter Thorburn, CSIRO, Australia
7. Crop Modeling Team Goal
• To evaluate different crop models
– for accuracy of response to climatic, CO2, and
other growth and management factors
– so there is confidence in the ability of models to
predict global change effects and make consistent
scenario-based projections of future crop
production for economic analysis.
Learn from intercomparisons and improve the
crop models. 2nd I in AgMIP is “Improvement”.
8. Crop Modeling Team Activities
• Activity 1 – Inter-compare crop models for methods and
accuracy of predicting response to variety of drivers
• Activity 2 – Conduct uncertainty pilot analyses across an
ensemble of models
• Want standardized protocols across crops.
– Wheat “uncertainty” (Asseng, Ewert)*
– Maize “uncertainty” (Bassu, Durand, Lizaso, Boote)*
– Sugarcane “uncertainty” (Thorburn, Marin, Singels)*
– Rice “uncertainty” (Bouman, Tao, Hasegawa, Zhu, Singh, Yin)*
– New teams (sorghum (Rao), peanut (Singh), potato (Quiroz))
*Already at work
9. Accomplishments Crop Modeling Team
AgMIP-South America Workshop
• Calibrated for two Brazilian sites
– three maize models (CERES-Maize, APSIM, &
STICS)
– two rice models (APSIM-ORZYA, and CERES-Rice)
• accounting for soils, cultivar, & management
• Used time-series and end-of-season data
10. Accomplishments Crop Modeling Team
AgMIP-South America Workshop
• Conducted climate change uncertainty analyses with
three maize and two rice calibrated crop models
– Mean temperature (Tmax & Tmin), (-3, 0, +3, + 6, +9 C).
– CO2 levels (360, 450, 540, 630, & 720 ppm)
– Rainfall (-30, 0, +30%)
– N fertilizer (0, 25, 50, 100, 150% of reference N)
• Simulated baseline 30 years and one future scenario!
• Compare how crop biomass, LAI, grain yield, grain
number, N accumulation, seasonal T and E respond to
these factors across the different crop models.
11. Sensitivity analyses examples from AgMIP Workshop, Campinas, Brazil,
August 2011
Grain Yield and Biomass Response of DSSAT, APSIM, & STIC maize models to
temperature
CERES
STICS
APSIM
11
12. Sensitivity analyses examples from AgMIP Workshop, Campinas, Brazil,
August 2011
Days to maturity and ET of DSSAT, APSIM, & STIC maize models in response to temperature. ET
affected by life cycle.
12
13. Sensitivity analyses Yield Yield
examples from
1.5
1.5
AgMIP Workshop a) b)
Campinas, Brazil
Relative Yield
Relative Yield
1.0
1.0
August 2011
0.5
0.5
APSIM APSIM
Upland Rice Upland Rice
0.0
0.0
DSSAT DSSAT
-2 0 2 4 6 8 400 500 600 700
Yield Response of
Temperatura CO2 level
APSIM-ORYZA
and CERES-Rice
to temperature, Yield Yield
CO2, rainfall, and
1.5
1.5
c) d)
N fertilization
Relative Yield
Relative Yield
Alex Heinemann,
1.0
1.0
Brazil, Aug 2011
0.5
0.5
APSIM APSIM
Upland Rice APSIM
Upland Rice
0.0
0.0
DSSAT DSSAT
CERES -30 -20 -10 0 10 20 30 0 50 100 150
Precipitation Variation N Levels
14. DOYMaturity DOYMaturity
1.5
1.5
Maturity Response r) s)
of APSIM-ORYZA
Relative DOYMaturity
Relative DOYMaturity
and CERES-Rice
1.0
1.0
to temperature,
CO2, rainfall, and
0.5
0.5
N fertilization
Alex Heinemann, APSIM
Upland Rice APSIM
Upland Rice
0.0
0.0
DSSAT DSSAT
Brazil, Aug 2011
-2 0 2 4 6 8 400 500 600 700
Temperatura CO2 level
APSIM
CERES DOYMaturity DOYMaturity
1.5
1.5
t) u)
Relative DOYMaturity
Relative DOYMaturity
1.0
1.0
0.5
0.5
APSIM APSIM
Upland Rice Upland Rice
0.0
0.0
DSSAT DSSAT
-30 -20 -10 0 10 20 30 0 50 100 150
Precipitation Variation N Levels
15. BIOMASS BIOMASS
1.5
1.5
Biomass Response i) j)
of APSIM-ORYZA
Relative BIOMASS
Relative BIOMASS
and CERES-Rice to
1.0
1.0
temperature, CO2,
rainfall, and N
0.5
0.5
fertilization
Alex Heinemann, APSIM
Upland Rice APSIM
Upland Rice
0.0
0.0
DSSAT DSSAT
Brazil, Aug 2011
-2 0 2 4 6 8 400 500 600 700
Temperatura CO2 level
APSIM
CERES BIOMASS BIOMASS
1.5
1.5
k) l)
Relative BIOMASS
Relative BIOMASS
1.0
1.0
0.5
0.5
APSIM APSIM
Upland Rice Upland Rice
0.0
0.0
DSSAT DSSAT
-30 -20 -10 0 10 20 30 0 50 100 150
Precipitation Variation N Levels
16. LAI LAI
1.5
1.5
LAI Response of e) f)
APSIM-ORYZA
and CERES-Rice
1.0
1.0
Relative LAI
Relative LAI
to temperature,
CO2, rainfall, and
0.5
0.5
N fertilization
Alex Heinemann, APSIM
Upland Rice APSIM
Upland Rice
0.0
0.0
DSSAT DSSAT
Brazil, Aug 2011
-2 0 2 4 6 8 400 500 600 700
Temperatura CO2 level
APSIM
CERES LAI LAI
1.5
1.5
APSIM
g) h) DSSAT
1.0
1.0
Relative LAI
Relative LAI
0.5
0.5
APSIM
Upland Rice Upland Rice
0.0
0.0
DSSAT
-30 -20 -10 0 10 20 30 0 50 100 150
Precipitation Variation N Levels
17. Maize Crop Pilot – Preliminary Results
Simona Bassu, Jean Louis Durand,
Jon Lizaso, Ken Boote
Baron Christian, Basso Bruno, Boogard Hendrik, Cassman Ken, Delphine
Deryng, De Sanctis Giacomo, Izaurralde Cesar, Jongschaap Raymond,
Kemaniam Armen, Kersebaum Christian, Kumar Naresh, Mueller Christoph,
Nendel Claas, Priesack Eckart, Sau Federico, Tao Fulu, Timlin Dennis,
Jerry Hatfield, Marc Corbeels
18. Model Behaviour: Maize Crop Pilot
Preliminary Sensitivity Analysis
Low input information
….Response to Temperature (6 models)
Morogoro (Tanzania) Ames (Us)
1,8
1,8
1,6
1,6
1,4 1,4
1,2 1,2
Yield ratio
yield ratio
1 1
0,8 0,8
0,6 0,6
0,4
0,4
0,2
0,2
0
0
-5 0 5 10
-5 0 5 10
Temperature increase (°C) T increase
19. Models Behaviour: Maize Crop Pilot
Preliminary Sensitivity Analysis
Low input information
….Response to CO2 (6 models)
Morogoro (Tanzania) Ames (US)
1,5
1,5
1,4
1,4
1,3 1,3
yield ratio
Yield ratio
1,2 1,2
1,1 1,1
1
1
0,9
0,9
300 400 500 600 700 800
300 400 500 600 700 800
[CO2] ppm [CO2] ppm
20. AgMIP Initiatives – Track 1
Experimenters & Crop Modelers Workshops
Model Improvement
Track 1
Track 2
− Test against observed data on response to CO2, Temperature,
including Interactions with Water, and Nitrogen Availability
20
21. Calibration of CERES and APSIM maize
models against 4 seasons at Wa, Ghana
5000
Simulated versus observed maize yield at Wa, Ghana over 4
years, using CERES-Maize (data courtesy, Jesse Naab)
4000
Simulated Grain Yield, kg/ha
y = 0.833 x + 361
3000 R2 = 0.925
2000
1000
0
0 1000 2000 3000 4000 5000
Observed Grain Yield, kg/ha
22. Tested CROPGRO-Peanut model response to temperature.
Crop grown at 350 ppm CO2. Model mimics observed pattern of
biomass & pod mass vs. temperature with pod failure at 39C.
12000 AgMIP, test accuracy of
multiple crop models
Crop or Pod, kg / ha
10000 against data like this.
Arrow is Southern
8000
US crop cycle temp.
6000
4000 Sim - Pod
Obs - Pod
Sim - Crop
2000 Obs - Crop
0
25 30 35 40 45
Mean Temperature, °C Genetic Impr.
Heat tolerance
23. 4000
Predicted - 700
Seed Yield, kg / ha
Observed - 700
Simulated Seed 3000
Yield of Dry Bean
Montcalm vs. 2000
Temperature
1000
No change needed
in temp effect on 0
podset or sd growth 20 25 30 35 40
Mean Temperature, °C
6000
Final Biomass of Mod Sim
Crop or Pod, kg / ha
Obs - Crop
Dry Bean Montcalm Default Sim
4000
vs. Temperature
Made leaf Ps less
sensitive to high 2000
temperature
0
20 25 30 35 40
Mean Temperature, °C
25. Regional Modeling: Motivation
• Research -- and common sense! -- suggest that poor agricultural
households are among the most vulnerable to climate change and
face some of the greatest adaptation challenges
• Rural households and agricultural systems are heterogeneous,
implying CC impacts – and value of adaptation strategies -- will vary
within these populations
• Farmers’ choice among adaptation options involves self-selection that
must be taken into account for accurate representation of adaptation
options
• Impacts of climate change and adaptation depend critically on future
technologies and socio-economic conditions
• Goal of AgMIP regional modeling is to advance CC impact and
adaptation research through the development of Protocols for
systematic implementation of impact and adaptation analysis, inter-
comparison and improvement.
26. Regional Modeling Activities
• Regional SSA and SA Teams
– All teams use at least one standard modeling approach (TOA-MD and others
according to region, team composition and interests)
– All teams develop RAPs, adaptation scenarios for their regions, consistent with
global RCPs, SSPs and RAPs
– Further refine RAPs concepts and protocols
• Linking regional models to national/global models
– Methods for coupling global model prices, other variables to regional analysis
– Inter-comparison of global and regional model outputs?
• Linking climate data, crop & livestock models to regional economic
models
– Developing improved methods for systematic use of climate data, soils and other
biological data with crop & livestock models to characterize spatial and temporal
distributions of productivity for use with economic models
• Methods to assess uncertainty in parameters, model structure
– Parameter estimation methods based on survey, experimental, modeled and
expert data; functional form and distributional assumptions
– Within and between individual model levels (climate, crop, econ)
26
27. Example: New Methods for Linking Crop and
Regional Economic Models
• Question: how to quantify the future productivity of ag systems for
impact assessment and adaptation analysis, accounting for spatial
heterogeneity?
• Answer: use crop models to simulate relative yield distributions:
– y2 = (1+ /y1) y1 = r y1 giving r = (1+ /y1) where r = r + r , (0,1)
– Using this model, with observations on one system and plausible bounds on r &
r we can approximate mean, variance and between-system correlations for the
other system
– data for r & r can come from crop model simulations
Example: maize relative yield
distribution in Machakos,
Kenya
R = future yield/present yield
28. Sensitivity analysis of alternative methods of estimating
relative yield distribution with matched and unmatched site-
specific data and averaged data (simulated CC gains and
losses, using TOA-MD model for Machakos, Kenya)
100000
Analysis shows critical role that
80000
estimation of spatial variance
60000
(heterogeneity) plays in estimation
of distributional impacts.
40000
20000
Losses
0
0 10 20 30 40 50 60 70 80 90 100
-20000
1a = time-averaged, matched bio-phys & econ data by site
-40000
1b = matched bio-phys & econ data by site (not time averaged)
2a = time-averaged, unmatched bio-phys & econ data by site
-60000
2b = unmatched bio-phys & econ data by site (not time averaged)
3a = site-specific bio-phys data, spatially averaged econ data with
-80000
approximated spatial variance
5a = averaged bio-phys and econ data
-100000
5b = averaged bio-phys and econ data, approximated variance of bio-
phys data only
Percent of Farms
1a 1b 2a 2b 3a 5a 5b
28
29. Example: Using TOA-MD and RAPs to simulate
distributional impacts of CC and adaptation
strategies using dual-purpose sweet potato, Vihiga
and Machakos Districts, Kenya
(note effect of RAPs on base and estimated impacts)
Vihiga Machakos
Poverty Rate (% of farm population living on <$1 per day)
Scenario No Dairy Dairy Total No Dairy Dairy Irrigated Total
base 85 38 62 85 43 54 73
CC 89 49 69 89 51 57 78
imz 87 42 65 85 44 50 73
dpsplw 88 42 66 85 44 50 73
dpsp 85 41 63 83 43 50 71
dpsp1 85 36 60 83 41 49 71
dpsp12 85 30 58 83 38 48 70
RAP1 base 65 17 41 72 30 46 60
RAP1 CC 71 18 44 77 33 47 64
RAP1 imz 66 15 41 70 27 40 58
RAP1 dpsp 65 15 40 69 27 40 57
Source: Claessens et al. Agricultural Systems in press 2012
29
31. Why bother? We all have lots to do!
It matters
• Policy makers care if we tell them
Agricultural land use will expand dramatically
Agricultural prices will increase by 100% between
now and 2050
Climate change will increase the number of
malnourished children by 25%
Increased agricultural research expenditures can cut
both of those numbers in half
Policy makers want 1 handed economists
32. WHAT DO THE MODELS SAY
ABOUT AGRICULTURAL PRICES?
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
33. IMPACT: Economy, demography and
climate changes increase prices
(price increase (%), 2010 – 2050, Baseline economy and demography)
Minimum and maximum
effect from four climate
scenarios
Page 33
34. Alternate Perspectives on Price
Scenarios (perfect mitigation), 2004-
2050
IMPACT has
substantially greater
price increases
Page 34
35. Alternate perspectives on agricultural area
changes, 2004-2050
IMPACT has
IMPACT has land use increases in
negative net land
some countries and decreases
use change
elsewhere
Page 35
36. Activities
Phase 1, Single reference scenario
• Single set of common drivers – income,
population, agricultural productivity without
climate change
• What do models say about key outputs?
• Why do they differ?
Phase 2, Explore relevant scenario spaces
• E.g., RAPs as drivers
• Linkages to crop and regional economic models