1. Jeroen Groot, 26 March 2012
Quantitative trade-offs analysis
in agricultural systems
Fields, farms and territories
Pablo Tittonell
Farming Systems Ecology – Wageningen University, The Netherlands
Pablo.tittonell@wur.nl
www.facebook.com/FSE.WageningenUR
Analysis of Trade-offs in Agricultural Systems
Wageningen
19 February 2013

2. Outline
1. What are trade-offs?
2. How to quantify them?
3. Examples
i. Measurements and data
ii. Output of a dynamic household model
iii. Pareto optimisation through evolutionary design
iv. Inverse dynamic modelling (global search alg.)
v. Agent-based systems and games

3. What are trade-offs?
Situations in which two or more competing/ conflicting objectives
must be simultaneously satisfied to a certain degree
Objective B
B1”
Complementarity
B1
Substitution
B1’
Objective A
A1
Tittonell (2013) Chapter on Trade-offs evaluation, CIALCA Conf., Earthscan, in press.

4. Tradeoffs analysis
Objective B
B1
A1 A2 A3 Objective A

5. Services écosystemiques: biodiversité et séquestration de C
A Vihiga B Siaya
Aboveground C stock (Mg ha-1)
40 40
homegarden
annual crop
permanent crop
30 30
pasture
A) Trees 20
B) Hedgerows 20
40 20
Delta C stock (Mg farm-1)
Vihiga 10 Vihiga 10
Siaya l Siaya
ntia
30 p ote 0
15 0
tion 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5
stra
que
C-se C 10 Vihiga D Siaya
C-sequestration potential
20
Aboveground C density (kg m-2)
8 8
Windrow
Individual tree
Woodlot
6 6
10 5
4 4
0 0
0 5 10 15 2
20 0 5 10
2 15 20
it
wt
Current aboveground C stock (Mg farm-1)
0 0
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5
g
Homegarden index
Shannon
b lh Food crop
hh
wlt mh Pasture
t
e
Cash crop
Slop
Woodlot
Henry et al. (2009), Agriculture Ecosystems and Environment 129

6. Quantifying trade-offs Absolute change Relative change
ΔB” ΔB” ΔB’
< ΔB’
ΔA ΔA B0” B0’
Objective B <
ΔA ΔA
B1”- B0” < B1’- B0’
A0 A0
B0” ΔA ΔA
ΔB”
B1”
Complementarity
B0’
ΔB’ Substitution
B1’
Objective A
A0 A1
ΔA
Opportunity costs, shadow prices, payment for environmental services, etc.
Tittonell (2013) Chapter on Trade-offs evaluation,relative sensitivity, preference
Elasticity of substitution, partial CIALCA Conf., Earthscan, in press. rate, etc.

7. Mapping trade-offs
Objective: 400
Alternative I
Increase
350
incomes
300 Alternative II
Gross margin ($ ha-1)
250
Complementarity
Alternative III
200 Current
150
100
Substitution
50
Alternative IV
0
20 25 30 35 40 45
Objective:
Maintain soil Soil organic matter (t ha-1)
Modelling: fertility
• To generate ‘clouds’ of alternative solutions
• To delineate ‘frontiers’ of possibilities Management strategies
Objective Indicator Current Alternative I Alternative II
To maintain Soil organic matter (t ha-1) 40 28 36
soil fertility
To increase Gross margin ($ ha-1) 180 360 280
net incomes
Cost of maintaining soil C: 15 $ t-1 25 $ t-1
Tittonell (2013) Chapter on Trade-offs evaluation, CIALCA Conf., Earthscan, in press.

8. Quantitative trade-offs analysis: methods
1. Ad-hoc analysis
1.1 - By looking at data
1.2 - By formalising a problem (discussion, expert knowledge, etc.)
1.3 - By looking at the output of a dynamic model
2. Multi-objective ‘compromising’ using models
2.1 - Using optimisation models (e.g., linear programming)
2.2 - Using search algorithms (e.g., inverse modelling)
2.3 – Agent based-systems
Tittonell (2013) Chapter on Trade-offs evaluation, CIALCA Conf., Earthscan, in press.

9. Ad hoc analysis

10. (i) Trade-offs analysis: data
Biomass allocation at village scale
Burkina Faso ( Andrieu et al., subm.)
Demand for residues across regions
Conservation field/farm scale
Trade-offs at agriculture
• No-till
3
India-1 High
• Rotation Bangladesh
• Soil cover
2
Kenya
tlu ha-1
Ethiopia-2
Ethiopia-1 Medium
India-2
Smallholder
1
Zimbabwe
Niger-2
agriculture: crop
Mozambique Nigeria
Malawi Low
residues are in high
Niger-1
0
0 demand
2 4 6 8 10 12
Naudin et al. ha-1
persons 2011
Valbuena et al. 2012

11. Semi-quantitative trade-offs analysis
Fuzzy cognitive mapping
Lopez-Ridaura, 2002
Management INCOME EROSION
BIO-
LEACHING
INCOME
DIVERSITY VARIABILITY
alternatives
Input-intensive
Organic farming
Integrated farming
Traditional practices
LOW
MEDIUM
HIGH Kok, 2008

12. Napier grass production (t farm-1
Napier grass
FArm-scale Resource Management SIMulator
(ii) Dynamic household model
Maize
Maize production (t farm-1)
60
8 NUANCES
CROP FIELD SOIL 50
Potential, water- and Soil C dynamics CLIMATE
nutrient-limited yields
6 Water, N, P and K 40
Actual variability
Weed competition availability Scenarios
30
4 MARKET
HOUSEHOLD Factors
Sweet potato
field 1 Maize field 3 Objectives & decisions Products 20
(0.18 ha) (0.24 ha) 2 Investment, allocation
and expenditure 10
COMMONLAND
Labour availability Rangeland
0 Woodlots 0
LIVSIM
20 40 HEAPSIM
60 80 100 120
Napier grass Feed supply and OFF-FARM
field 2
(0.15)
demand
Productivity Soil Manure & qualityand-1Napier
B of storage collection, ha )
organic C (t
Maize Employment
Milk, meat, traction Remittances
and manure
1 1
Effects on soil fertility FARMSIM
Relative Napier grass yield
Maize field 2 0.8 0.8
(0.25 ha) A
Relative maize yield
10 70
Napier grass production (t farm-1)
Napier grass Napier grass production
0.6 Maize 60 0.6
Maize production (t farm-1)
8
Maize
Napier grass
field 1
Manure 50
0.4 0.4
field 1
(0.06 ha)
(0.15 ha) allocation 6
40
strategies
0.2 0.2
(10 year 4 Maize production 30
Manure
heap simulations) 20
0 0
2 1 2 3 4 5 6 7 8 9 10
10
Even spread Concentration
Homestead
2 cows 0 Manure allocation strategy 0
20 40 60 80 100 120
Rowe et al., 2006; Titttonell et al., 2007; 2009; Van Wijk et al., 2009; Rufino et al., 2011; )Zingore et al., 2011
Soil organic C (t ha-1

13. Multi-objective ‘compromising’

14. (iii) Pareto optimisation: evolutionary design
Evolutionary model
generate
by allocating
land-use activities
evolutionary algorithm
Obj. 1
evaluate Pareto-ranking
for multiple
indicators
1
2 1
2 1
2 1
2
rank & select
using non-weighting
dominated
Pareto-based methods
Farm IMAGES Obj. 2
Landscape IMAGES
Groot & Rossing (2011)

15. (iii) Pareto optimisation: evolutionary design
Evolutionary model
generate
by allocating
land-use activities
evolutionary algorithm
Obj. 1
evaluate Evolutionary algorithm
for multiple
indicators
rank & select
using non-weighting
Pareto-based methods
Farm IMAGES Obj. 2
Landscape IMAGES
Groot & Rossing (2011)

16. Intensification pathways at farm scale
Productivity per animal
a. Trade-offs
2400
2000
Labor balance (h)
Groot et al., 2012. Agricultural Systems.
1600
Productivity per unit labour 1200
800
400
0
-20 0 20 40 60 80
500 500
Organic matter balance (kg/ha)
b. c.
250 250
0 0
-250 -250
-500 -500
-20 0 20 40 60 80 0 400 800 1200 1600 2000 2400
80 80 80
d. e. f.
Soil nitrogen loss (kg/ha)
70 70 70
60 60 60
50 50 50
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
-20 0 20 40 60 80 0 400 800 1200 1600 2000 2400 -500 -250 0 250 500
Cortez-Arriola et al., subm. Operating profit (k€)
Ecological services Ecological services (h)
Labor balance Organic matter balance (kg/ha)

17. (iv) Inverse modelling
• A spatially heterogeneous farm Trade-offs between objectives
200
• A limited availability of cash 25
180 10000 KSh
• A limited availability of labour
Farm farm scale (kg)
24 5000 KSh
2000 KSh
• Objectives: maximise food 23
Relative investment in erosion control erosionfarm scale (t)
160
Farm erosion at loss [tons]
N losses at N loss [kg]
production, minimise N 22
losses, etc… 21
140
Simulated management decisions
20
A 120
B
Soil soil
0.8
Profile 0.8
19
Relative investment in weeding
2000 KSh 100
2000 KSh
Homestead
18
Napier grass
5000 KSh 5000 KSh
Compound
0.6 10000 KSh 0.6
17 10000 KSh
fields Home garden 80
0 1000
1 2000
2 3000
3 4000
4 5000
5 6000
6 7000
7 8000
8
Maize fields
Living fence Woodlot 16 Farm grain yield [kg]
0 1000 2000 3000 4000 5000
Tea 0 1 Farm-scale3maize4grain 5
2 production (tones)8000
6000
6 7000
7 8
0.4 Maize
0.4 Farm grain yield [kg]
Farm-scale maize grain production (t)
Layout
0.2 0.2
Maize 1 Sweet
Maize 2 potato Maize 5
(+) (-)
(+/-)
0 Maize 6 Woodlot
0.0
Maize 4 (-)
0 0.2 (+/-)
0.4 Tea 0.6 0.8 0.0 0.2 0.4 0.6 0.8 1.0
Maize 3
Home Relative investment in N fertiliser
(+) RelativeTittonell et al. (2007), Agricultural Systems 95
investment in land preparation
garden

18. Agent-based systems

19. (v) Agent-based systems
Multi-scale –trade-offs around crop residue biomass herd
A village territory representation of the multi-agent modeltypes, communal
Figure 2 Schematic of 100 Km2, 4 farm
use in the Zambezi valley
Results at village scale
Baudron, Delmotte, Herrera, Corbeels, Tittonell
5.5 0
Average change of total soil organic carbon
5 0 kg N ha-1 (a)
Intensification through conservation agriculture to preserve habitats and biodiversity
-2 0 kg N ha-1 (b)
4.5 20 kg N ha-1 20 kg N ha-1
-4
Average mulch cover (t ha-1)
4 100 kg N ha-1 100 kg N ha-1
in the 0-20 cm ( t ha-1)
-6
3.5
3 -8
2.5 -10
2 -12
1.5
-14
1
0.5 -16
0 -18
0 100 200 300 400 0 100 200 300 400
Cattle density (head km-2) Cattle density (heads km-2)

20. Simulation and gaming - Mexico
Mapa de la Reserva de la Biosfera de la Sepultura. Fuente: CONANP
Simulation and gaming for improving local adaptive
capacity;
The case of a buffer-zone community in Mexico
E.N. Speelman (2008-2013)
Supervisory team
J.C.J. Groot, L.E. Garcia-Barrios, P. Tittonell

21. A methodological framework Landscapes
COMPASS
Attic LandscapeIMAGES
ActorIMAGES
Agro-ecosystem diversity, Trajectories and Trade-offs for Intensification of Cereal-based systems
Economic Spatial Land use
results coherence systems
Farms
Nutrient Landscape Collective
losses metrics decisions
Diego Valbuena (WUR)
Bruno Gerard (CIMMYT)
Nutrient Jeroen Groot (WUR)
Water Feed FarmIMAGES
balance balance balance
Fields, landscape elements
Santiago Lopez Ridaura (CIMMYT)
FarmDESIGN
FarmSTEPS
Labor
balance
Fred Baudron (CIMMYT)
Economic
results
Nutrient
losses FarmDANCES
Andy McDonald (CIMMYT)
Tim Krupnik (CIMMYT)
Felix Bianchi (WUR)
Katrien Descheemaker (WUR)
Nutrient Organic Soil Water FieldIMAGES Pablo Tittonell (WUR)
balance matter erosion balance
NDICEA
Crop yield
Nutrient Nutrient Plant RotSOM
ROTAT
3 PhD started in 2013
uptake losses diversity
A Cimmyt-Wageningen collaboration in the context ofSimulation – Groot and Wheat
Co-innovation and Modeling Platform for Agro-ecoSystem the CRP Maize et al., 2012

22. Summary
Trade-offs: situations in which two or more competing/ conflicting
objectives must be simultaneously satisfied to a certain degree
Quantifying slopes, opportunity costs or substitution rates not always
enough – models can be used to map-out tradeoffs, to explore a wider
range of options and possibility frontiers
Model-aided trade-offs analysis:
1. Dynamic household models (no formal optimisation)
2. Optimisation through linear programming
3. Pareto optimisation through evolutionary algorithms
4. Agent-based systems
How to scale? Models typically work for single ‘representative’ farms;
typologies, distribution of farm population, etc.
How to choose? Objective algortithms can always be calculated, but
they cannot replace the insihgt to be gained by involving the actor;
combinations of both aproaches are possible