Session 5.2 Combining numerical modeling with a participative approach
1. Combining numerical modeling with a
participative approach
for the design of sustainable and applicable cropping systems
Meylan, L., Sibelet, N., Gary, C., Rapidel, B.
1
2. Scientific context
• Objective: designing agroforestry systems with
multiple performance requirements
• Challenge: interaction of several spp; complex
and long-term effects to take into account
• Two common approaches
?
– Modeling: useful for complex interactions often
found in agroforestry; many trials, low cost; but
theoretical results and few developed models
Vereijken, 1997; Rapidel et al, 2009; Whitbread et al, 2009
– Participative: involving farmers, suitable
solutions, but solutions not generic
Tixier et al, 2006; Malézieux et al, 2009
3. Case study: shaded coffee in
Llano Bonito, Costa Rica
• Coffea arabica (mostly Caturra) + shade trees (mostly
Erythrina and Musa spp)
• Relatively small farms (1-2ha) supporting many families;
600 farmers in the Llano Bonito watershed (18km²)
• Intensive production (400kg of N/ha/yr) sold at
premium price
4. Case study: shaded coffee in
Llano Bonito, Costa Rica
• High erosion rates due to steep
slopes and rainy season
(+3000 mm of rain/year)
• Operation of downstream
hydroelectric dam threatened by
sedimentation of Pirris river
• Interest in PES (Payment for
Environmental Service) scheme
for coffee growers to reduce
erosion
5. Case study: shaded coffee in
Llano Bonito, Costa Rica
Coffee production
shade trees
erosion
• Trees help prevent erosion
• … but how many trees in coffee systems to
maximize performance in both areas?
?
Actual shade tree densities vary
between 100 and 400 trees/ha;
most trees are pruned 2-3 times a year
Shade trees can have positive or negative
effects on coffee production depending on
climate, environment, management…
Minimum shade level needed – what is the optimum?
7. CAF: numerical model for coffee
shaded systems
Can we combine the information from a
numerical model with farmers’ knowledge to
design sustainable AND practicable systems?
Van Oijen et al, 2010
8. Research objectives
1. Involve coffee growers in the modeling
process in order to identify potential systems
that farmers would be willing put into place
themselves
2. Evaluate usefulness of CAF2007 in generating
improved agroforestry systems that are
acceptable to farmers
9. Participative modeling sessions
• Preparation
– Previous data and conceptual
model to introduce trees, coffee
& erosion relationship
– Progressively introduce new
concepts – graphs/charts, model
• Participants asked to think about what they
would test “on the field”
• All questions and suggestions recorded verbatim
– number of variables in each “what if” question
– topics covered
10. Participative modeling sessions
• 19 farmers in 4 groups; 5 sessions of 2hrs each
S1
S2
S3
S4
S5
• Ideas for experimentation – no model
• Conceptual model, check outcomes of S1
• Numerical data
• What does “model” mean?
• Presentation of CAF2007, potential scenarios to be simulated
• Evaluation of simulation outcomes
11. Participative modeling sessions
6
15
Low intensive
Work intensive
5
Shaded systems
Complexity
(mean number
of variables per
question)
Diversity
(mean number
of variables per
farmer)
Inputs intensive
4
3
10
Low intensive
5
2
Work intensive
Shaded systems
1
Inputs intensive
0
0
S1
S2
S4
Working session
S5
Increased
participation, complexity of
questions, and number of topics
covered after introduction of
CAF2007 in S4
S1
S2
S4
Working session
S5
12. Evaluation of CAF2007 by participants
Mostly disagreed
with model output
•
•
•
Shade tree density
Shade tree pruning
Coffee pruning
• High interest in simulation of
N cycling and fertilization
• Model inaccuracies: effect of
shade tree on yield over
several years
Missing
•
•
Pests & disease
Annual yield
variations (e.g.
dieback effect)
Mostly agreed with
model output
•
•
•
Fertilization
Mineral N pool
Coffee LAI and
vegetative growth
13. Conclusions
• Coffee growers actively engaged with a
numerical model to discuss changes in their
management
– “what if?” and simulation requests; increased
number, scope and complexity with CAF2007
– both positive and negative feedback on model
performance
– exploration of normally “invisible” processes shown
in model, e.g. erosion, mineral N pool
• New systems (change in management) were
identified – to be tested by growers themselves
17. Typology of coffee plots
1: low
intensity
2: labor
intensive
3: shaded
systems
4: agrochemical
intensive
Fertilizer
(USD/ha/yr)
518 c
1395 a
909 b
1157 a,b
Fungicide
(USD/ha/yr)
51 a,b
28 b
16 b
79 a
Manual weed
control
(hrs/ha/yr)
131 a
203 a
120 a
144 a
Shade tree
density (#/ha)
288 a,b
332 a,b
539 a
235 b
Yield (t/ha/yr)
4.2 b
8.9 a
7.2 a
8.1 a
1.1 a,b
1.9 a
0.6 b
1.2 a,b
Erosion control
18. Effect of shade tree density
on coffee yield
Coffee yield (t/ha/yr)
12
10
R² = 0.301
Group 1
8
Group 2
6
Group 3
R² = 0.565
Group 4
4
2
0
200
400
600
Shade tree density
800
1000
21. Effect of shade trees on coffee
1.0
5
0.9
Infiltration delay 0-30cm (h)
5
kg of litter per m²
4
4
3
3
2
2
1
1
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
full sun
Erythrina
shade treatment
0.0
banana
Full sun
Erythrina
Banana
6000
g of dry vegetation/m²
5000
4000
Full sun
3000
Erythrina
2000
Banana
1000
0
0.40
0.50
0.60
0.70
0.80
0.90
Infiltration delay (h)
1.00
1.10
1.20
No significant effect
of shade on
production; but
positive effect on soil
litter and
infiltration, used
here as proxies for
erosion (actual runoff and
erosion data forthcoming)
22. Preparing the model
Variable
Yield
Site 1
Site 2
Site 3
Initial simulation
1.22
1.68
1.69
2.01
1.2
0.52
0.88
0.67
Initial simulation
1.22
1.48
1.63
After calibration
0.89
0.70
Initial simulation
2.30
2.01
1.87
0.34
0.46
0.33
Initial simulation
0.22
-
0.14
After calibration
0.06
-
Left: RMSE for
output values
used in model
calibration
0.45
After calibration
Soil water content
1.08
After calibration
Coffee wood biomass
3.47
Initial simulation
Tree LAI
1.95
After calibration
Coffee LAI
2.36
0.04
Carbon in coffee wood
1
0.8
kg of C / m2
Right: example of
calibration effect on
simulated coffee
wood biomass values
after
calibration
0.6
measured
data
0.4
0.2
before
calibration
0
4000
4200
4400
4600
days of simulation
4800