Manual of FALLOW Model
Version 2.1
R. Mulia, B. Lusiana, and D. A. Suyamto
World Agroforestry Centre
2013
Disclaimer and Copyright
This is a model on rural landscape dynamic. Although efforts have
been made to incorporate relevant process knowledge on a range of
interactions, the model is not more (and not less) than a research
tool. Model prospective outputs may help in developing specific
hypotheses for research, in exploring potential future options on
development strategies, but they should not be used as authoritative
statements per se.
Copy right, but do not copy wrong. The FALLOW model was
developed on the basis of publicly funded research at the World
Agroforestry Centre (ICRAF) and may be used for non-commercial
research purposes in the interest of research or governmental
institution as well as farmers of the world.
-------------------------------------------------------------------------------For further information, consultation, and technical support please
contact:
Betha Lusiana ([email protected]),
Rachmat Mulia ([email protected]), or
Meine van Noordwijk ([email protected])
World Agroforestry centre (ICRAF)
Southeast Asia Program
P.O. Box 161
Bogor 16001 Indonesia
Table of contents
1. Introduction
A. What is it?
B. How it works?
C. How to use it?
D. How to get the model?
2. Examples of model application
A. Which development strategy is most appropriate:
genuine or pseudo development?
B. S-curve story of subsidy for development
3. Description of FALLOW modeling concepts
A. Landuse option and change
B. Succession and growth
C. Land productivity
1
2
3
3
4
5
8
14
15
Appendix 1. List of input maps and variables for FALLOW
simulation
Appendix 2. Minimum inputs for FALLOW simulation
Appendix 3. Running FALLOW model with Nutshell
Appendix 4. Important notes in running FALLOW
18
26
28
31
References
35
1. Introduction
A. What is it?
Figure 1 Prospective diagram depicting the impact of
development strategies to economical (x axes) and
ecological value (y axes) relative to the initial
condition before implementing the strategies
(baseline, central point of the diagram).
The main issues in
prospecting landuse
strategies for rural
landscapes are mostly
related to:
• Non-linear
baseline
trajectories;
• Trade-offs
between
economical
utilities and
environmental
services; and
• Additionality
The FALLOW Model has been developed as a tool to prospect the
likely shifts of some scenarios on such strategies from the baseline.
The strategies may imply to:
•
•
•
•
Losses in both economical and ecological values (collapse);
Gains in economical value but loss in ecological value (red
development);
Gain in ecological value but loss in economical value
(conservation); or
Gains in both economical and ecological values (green
development).
1
B. How it works?
The FALLOW Model simulates
land use/cover change dynamics
due to local responses on external
drivers with various feedback
loops, and assess the
consequences of the resulting land use mosaics on economical
utilities (welfare and food security) and environmental services
(carbon stocks, watershed functions and biodiversity).
Local responses portrayed by the model comprise:
• How farmers adjust their expectation about economical utility of
each available option on land-based and non-land-based
investments through learning;
• How farmers allocate their capitals (labors, money and land) to
each available option of investments;
• How farmers perceive about attractiveness of a plot to expand
particular land use system, with regards to some spatial factors
determining potential benefits (soil fertility, suitability and
attainable yield) and potential costs (transportation,
maintenance and land clearing);
• Succession, growth, fire and land conversion; and
• Laws of diminishing and increasing marginal utilities on soil
fertility and land use productivity.
The main external drivers incorporated in the model include:
• Market mechanisms and relevant regulation interventions,
articulated through commodity prices, costs and harvesting labor
productivities;
• Development programs, articulated through extension, subsidies,
infrastructures (settlements, road, market, processing factories),
and land use productivities; and
2
•
Conservation programs, articulated through forest reserves as
prohibited zones for farmers.
C. How to use it?
The FALLOW Model is a raster-based spatially explicit model with
default spatial resolution of 1 ha, temporal resolution of 1 year and
socio-economical resolution of 1 community, applicable for rural
landscapes. The model uses PCRaster (http://pcraster.geo.uu.nl) as the
main platform. Your computer should have operating system of
Microsoft Windows XP Professional, version 2002 or later, which
processor having at least 3.2 GHz of speed and 496 MB of RAM, and
which hard disks having at least 15 GB of free space.
D. How to get the model?
We provide you with free sources of the model, which are freely
downloadable from: http://worldagroforestrycentre.org/sea/ or upon
request to [email protected]. We also provide you with educational
versions of the model, developed using STELLA and NetLogo.
3
2. Examples of model application
A. Which development strategy is most appropriate: genuine or
pseudo development?
Figure 2. FALLOW model as applied in 4 regions in Indonesia: Muara Sungkai (MS)
and Way Tenong (WT) in Lampung, Sumatra; Sindenreng Rappang (SR) in southern
Sulawesi; dan Sebuku (SB) in western Kalimantan.
Approach for stimulating the installation of tree-based landuse
systems in developing countries is usually of two types: project or
programmatic approach, and both involve some regulations and
interventions applied to farmer society.
Project-approach is usually characterized by top-down
incentive, applied only in the previously-determined project area and
duration without any substantial effort to solve the apparent
problems experienced by the society in the observed area.
Programmatic-approach has more intentions to overcome the
4
apparent problems which relate to e.g. access to market, to land, or
to source of knowledge or information.
Model applications in four regions in Indonesia (Figure 2)
show that programmatic-approach (blue circles) induced the
development of tree-based landuse systems by farmers and can likely
convert marginal lands with relatively low investment cost. This
offers better economical (x axes) and ecological (y axes) merit than
project-approach (red circle). Project-approach mostly induces
ecological benefit, but inferior related economical value. Please see
van Noordwijk et al. (2008) for a more detail description and
explanation.
B. S-curve story of subsidy for development
The model has also been applied to simulate the effect of subsidy
expected to accelerate the conversion of pioneer forest (i.e. marginal
lands with shrubs) to rubber-based landuse systems in Muara
Sungkai, Sumatra. Without any subsidy, it is unlikely that this region
could be converted into landscape with prevailing rubberplantations. The results suggest, however, that prolonged subsidy is
no longer efficient when land and labour availability are limiting
(Figure 3-6) (Suyamto et al., 2005).
For other examples of model application, please see Lusiana et al.
(2012), Mulia et al. (2013), and Tata et al. (2013).
5
Figure 3. Landcover dynamic without any subsidy for development of rubber-based
landuse system.
Figure 4. Limiting resources in the development of rubber-based landuse system
without subsidy.
6
Figure 5. Landcover dynamic with subsidy for the development of rubber-based
landuse system.
Figure 6. Limiting resources in the development of rubber-based landuse system
with subsidy.
7
3. Description of FALLOW modeling concepts
Description of FALLOW model principles can also be seen in van
Noordwijk (2002) and Suyamto et al. (2009). Below is a more detailed
description with equations.
A. Landuse option and change
In the current version, 15 livelihood options can be considered by
farmers: off-farm, NTFP, timber/logging, 4 types of agricultural crops,
and 8 types of agroforesty/plantation. Profit from each option is
calculated relative to land area or labor (i.e. payoff to land and to
labor respectively):
𝑃𝑖𝑙𝑎𝑛𝑑 =
(𝑦𝑖 ∗ℎ𝑖 )−𝑐𝑖
𝑃𝑖𝑙𝑎𝑏𝑜𝑟 =
𝑎𝑖
(𝑦𝑖 ∗ℎ𝑖 )−𝑐𝑖
𝑙𝑖
(1)
(2)
Where Pi is profit in unit currency person day-1 related to labor
and in unit currency unit area-1 related to land obtained with
livelihood option i, yi is total attainable yield in the year (unit yield), hi
is average yield price in the year (unit currency unit yield-1), ci is total
cost (unit currency), ai is total harvesting area (unit area), and li is
total labor involved (person day). Total cost c comprises of labor and
non-labor cost. The first accounts for external labor cost, i.e. salary
for labors coming from outside local community. Non-labor cost
consists of establishment and maintenance cost that are not related
to labor, e.g. cost for buying seeds and/or fertilization at planting
time or during productive season in the year. Any subsidy from
government for plot establishment or maintenance reduces the nonlabor cost. Total labor l is the total of allocated ‘internal’ labors to
livelihood option i (see equation 12 below) plus external labor
8
involved in that option. Hereafter, unit currency is represented in Rp,
unit area in ha, and unit yield in ton.
Expected payoff to land and labor
Farmers do have certain profit expectations for their livelihood
options. Next year profit expectation is determined based on
knowledge obtained in current year:
𝐸𝑡+1,𝑖 = 𝐸𝑡,𝑖 +∝ (𝑃𝑡,𝑖 − 𝐸𝑡,𝑖 )
(3)
Where Et,i is current year profit expectation for livelihood option
i and  reflects farmer’s adjustment rate to that expectation after
current year profit analysis. Equation 3 is used to calculate expected
payoff to land and to labor. The same  value is applied for the two
calculations. Farmer’s profit expectation, however, might not be in
accordance with that of outsiders, e.g. economists. ‘Non-steady’
farmers might consider suggestions from others for calculating their
‘final’ profit expectation:
𝑓𝑖𝑛𝑎𝑙
𝐸𝑡+1,𝑖 = 𝐸𝑡+1,𝑖 + 𝛽(𝑆𝑡+1,𝑖 − 𝐸𝑡+1,𝑖 )
(4)
Where St+1,I is profit expectation suggested by others and 
reflects farmer’s adjustment rate to their own expectation:
𝛽 = 𝑆𝑎𝑣 ∗ 𝑆𝑒𝑥𝑝 ∗ 𝑆𝑐𝑟𝑒𝑑
(5)
Sav is suggestion availability (0 or 1) e.g. through extension or
group discussion; Sexp is farmer’s exposure to attend the extension (01); and Scred is farmer’s assessment to extension credibility (0-1). A
similar value of  is used to calculate final expected payoff to land
and labor. Two types of farmer’s community can be simulated with
different  and/or  value.
9
Resource allocation based on profit expectation
Based on profit expectations, farmers determine variety and intensity
of their future livelihood options. In principle, productive plots will be
maintained whereas unproductive ones are possible to change into
another landuse type. Farmers might also open new plots for
profitable livelihood options. In the model, 3 types of landuse are
simulated: forest, agriculture, and agroforestry or plantation. Forest
consists of 4 classes based on aboveground biomass: pioneer, young
secondary, old secondary and primary forest. Agroforestry or
plantation also consists of 4 classes based on productivity level:
pioneer, early production, late production, and post production.
Potential areas for conversion are those of 4 types of forest except
protected forests, agroforestry/plantation plots that are at late or
post production stage, and agricultural crop plots with production
less than 0.5 ton ha-1. Suppose that Apot describes total area potential
for conversion (ha) and Apot,i is total area potential for conversion into
livelihood i:
𝑙𝑎𝑛𝑑
𝐴𝑝𝑜𝑡,𝑖 = 𝐴𝑝𝑜𝑡 ∗ 𝑓𝑡+1,𝑖
(6)
And:
(𝑤𝑖 ∗𝐸𝑡+1,𝑖 )𝜌
𝑙𝑎𝑛𝑑
𝑓𝑡+1,𝑖
= ∑𝑛
𝑖=1(𝑤𝑖 ∗𝐸𝑡+1,𝑖 )
𝜌
(7)
That basically describes relative expected payoff to land. The
power  reflects farmer’s degree of profit-orientation: >0 means
livelihood options with higher profit expectations are prioritized, =0
means equal prioritization and <0 gives more prioritization to less
profit options. It is also possible that farmers prefer certain
livelihoods option due to a non-economic reason such as cultural
reason. For this, different weighting values wi can be set between
10
livelihood options. In the current model version, farmers might be
different in  value but not in weighting factors.
Available labors should be distributed to existing productive plots
for maintenance and/or to potential converted areas for land
clearing. Total available labor (person day) is:
𝐿 = 𝐻 ∗ 𝑓𝑙𝑎𝑏𝑜𝑟 ∗ 𝑛𝑑𝑎𝑦𝑠
(8)
Where H is total population (person), flabor is labor fraction (0-1),
and ndays is annual working day (person day). Total labor for livelihood
option i is:
𝑙𝑎𝑏𝑜𝑟
𝐿𝑖 = 𝐿 ∗ 𝑓𝑡+1,𝑖
+ 𝐿𝑒𝑥𝑡,𝑖
(9)
Where Lext is total external labor required for livelihood option i
(person day) and flabor is calculated in the same way as fland in
equation 7 but with expected payoffs to labor instead to land. No
differentiation in  and w values are made between the two
calculations. No change in demand for external labor is assumed
throughout the years. If different types of farmer were simulated, a
fraction will determine population of each farmer’s type and labor
allocation to livelihood options (equation 9) will be carried out for
each group.
Total area potential for conversion into a specific livelihood
option has been calculated in equation 6 but not their specific
locations in the landscape. Probability that a certain plot within
potential areas for conversion is reserved for livelihood option i is
Apot,i/Apot. Suppose that Fi is cumulative probability for Apot,i/Apot then
any plot is reserved for option i if Fi-1<r<Fi where r is a generated
random number (0-1), F0 = 0 and F11 = 1. If size of a plot is m ha, there
are Apot/m plots to process.
11
Actual converted areas
Opening new plots needs labors for land clearing and budget for
covering establishment cost. All allocated labors calculated in
equation 9 can be considered as available labors for land clearing by
assuming they are firstly allocated to establish new plots and later to
maintain all existing including new established plots. Budget for
covering establishment cost can come from two sources: allocated
budget by farmers for opening new plots (explained below) and
subsidy from government. Total actual converted area (ha) is a
function of labor and financial capacity:
𝐿 𝑀𝑒𝑥𝑝,𝑖
, 𝐴𝑝𝑜𝑡,𝑖 )
𝑑𝑖 𝑐𝑒𝑠𝑡,𝑖
𝐴𝑎𝑐𝑡,𝑖 = min⁡( 𝑖 ,
(10)
Li is allocated labor (equation 9), di is labor requirement for plot
establishment (person day ha-1), Mexp,i is available budget for land
expansion (Rp, see equation 21 below) allocated to each livelihood
type proportional to labor allocation fraction (flabor, equation 9) plus
subsidy for establishment if any, and cest,i is establishment cost (Rp
ha-1).
Location of actual new plots
For each livelihood option, plot selection should be done if total
actual is less than total potential converted area (i.e. Aact,i < Apot,i).
The selection is based on ‘attractiveness index’:
𝑋𝑖,𝑗 =
𝑤𝑓𝑒𝑟𝑡,𝑖 ∗𝑓𝑒𝑟𝑡𝑗 +𝑤𝑠𝑢𝑖𝑡,𝑖 ∗𝑠𝑢𝑖𝑡𝑗 +𝑤𝑝𝑦𝑖𝑒𝑙𝑑,𝑖 ∗𝑝𝑦𝑖𝑒𝑙𝑑𝑗
1+𝑤𝑡𝑟𝑎𝑛𝑠,𝑖 ∗𝑡𝑟𝑎𝑛𝑠𝑗 +𝑤𝑚𝑎𝑖𝑛,𝑖 ∗𝑚𝑎𝑖𝑛𝑗 +𝑤𝑠𝑙𝑜𝑝𝑒,𝑖 ∗𝑠𝑙𝑜𝑝𝑒𝑗 +𝑤𝑓𝑏𝑖𝑜𝑚,𝑖 ∗𝑓𝑏𝑖𝑜𝑚𝑗
(11)
Xij is attractive index of plot j reserved for livelihood option i;
wfert,i describes importance of soil fertility factor for option i (0-1), e.g.
oil palm plantation that needs higher level of soil fertility than rubber
12
plantation has a higher wfert value. Fertj is actual level of soil fertility
represented qualitatively (1=very poor-5=very fertile); suitj is plot
suitability (0 or 1); pyieldj is potential yield (ton ha-1) (see explanation
below for potential and actual yield). Transj is distance (m) between
plot j and the closest transportation network:
𝑡𝑟𝑎𝑛𝑠𝑗 = min⁡(𝑟𝑜𝑎𝑑𝑗 , 𝑟𝑖𝑣𝑒𝑟𝑗 , 𝑚𝑎𝑟𝑘𝑒𝑡𝑗 , 𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑗 )
(12)
Roadj, riverj, marketj, and industryj are the closest distance (m)
from plot j to road, river, market, and processing industry
respectively. Mainj is the closest distance to maintenance centre (i.e.
settlement or existing plots with the same livelihood option):
𝑚𝑎𝑖𝑛𝑗 = min⁡(𝑠𝑒𝑡𝑗 , 𝑒𝑥𝑖𝑠𝑡𝑗 )
(13)
Slopej and fbiomj measure plot steepness (%) and floor biomass
(ton ha-1) respectively. Attractive index basically compares profit that
can be derived from factors supporting yield (soil fertility, land
suitability, and potential yield) and potential cost for accessing,
clearing and maintaining plot (transportation, maintenance, land
slope, and floor biomass). Based on their degree of attractiveness,
potential new plots are classified into 5 categories: z < 0 = very
unattractive plot, 0  z < 1 = less attractive plot, 1  z < 2 = quite
attractive plot, 2  z < 3 = attractive plot and z  3 = very attractive
plot where z is normally standardized attractive index.
Now suppose ni,k is number of potential new plots reserved for
option i that classified into attractiveness category k (k=1, 2, …, 5 for
very attractive) and gi,k describes number of plots with higher
categories than k. Thus, gi,5 = 0, gi,4 = ni,5 + gi,5, …, and gi,1 = ni,2 + gi,2.
Probability that a plot (between potential new plots for livelihood
option i) with attractiveness category k will be selected as actual new
plot is:
13
𝑝𝑖,𝑘 =
𝐴
min⁡(𝑛𝑖,𝑘 , 𝑎𝑐𝑡,𝑖 −𝑔𝑖,𝑘 )
𝑚
𝑛𝑖,𝑘
(14)
If a generated random number is less than pi,k then plot with
category k will be selected as actual new plot for livelihood option i.
m is size of a plot (ha). Equation 14 basically ignores potential new
plots with a lower attractiveness category if total number of plots
with a higher category is still sufficient to satisfy Aact,i/m. Farmers
might use fire for opening lands and fire can spread to adjacent plots
if a generated random number is less than the probability that fire
escapes from one plot to its adjacent plots.
B. Succession and growth
Initial plot age and aboveground biomass of each landcover system
are estimated based on a specified mean and variation between
plots. In the dynamic process, as long as plot is not converted into
another landuse type or damage due to fire accident or disaster, plot
age increases and plot can achieve a higher ecological maturity or
production status. Otherwise, age will be reset to 0. For new logged
forests, their ecological stage and age degrades depending on logging
intensity. In case farmers do slash and burn activity and fire escapes
to outside the opened plot then adjacent plots are converted into
pioneer forests. The opened plot however will be directly used for a
particular livelihood option. Plots affected by disaster and marginal
plots of any agricultural crop system that are not converted into
another livelihood option will also become pioneer forests; whereas
non-converted forests, late or post productive agroforestry or
plantation types will keep their status until selected as an actual new
plot later. Human population grows according to an annual
population growth rate but will decrease in case of a disaster.
Settlement growth is not simulated in the current model version.
14
C. Land productivity
Soil fertility
Soil fertility is defined as the ability of soil to support plant growth
and yield. In agricultural crop systems, biomass and yield are totally
harvested and not returned to the soil. This induces soil fertility
degradation since planting time. Soil recovery can occur due to e.g.
fertilization:
(𝑓𝑒𝑟𝑡𝑚𝑎𝑥⁡ –𝑓𝑒𝑟𝑡)2
1/2 )∗𝑓𝑒𝑟𝑡𝑚𝑎𝑥⁡ −𝑓𝑒𝑟𝑡
𝛿𝑟𝑒𝑐𝑜𝑣 = (1+ℎ
(15)
Where fertmax is maximum soil fertility reflecting inherent soil
fertility based on geological class, fert is actual soil fertility (1-5), and
h1/2 is half time recovery, i.e. time needed to achieve half of fertmax
(year). No soil degradation is assumed in other landuse types (i.e.
forest, agroforestry or plantation).
Harvest and yield
In agricultural crop systems, a unit decrease in soil fertility is
equivalent to a certain production level (ton ha-1). Crop yield is
therefore a function of soil depletion:
𝑝𝑜𝑡
𝑦𝑖,𝑗 = 𝑦𝑟𝑒𝑓,𝑖 ∗ 𝛿𝑑𝑒𝑝,𝑗
(16)
Where yref,i is attainable yield (ton ha-1) in plot of livelihood
option i with one unit decrease in soil fertility. dep,j is soil fertility
depletion in plot j which is not constant throughout the years:
𝛿𝑑𝑒𝑝,𝑡 = 𝑓𝑒𝑟𝑡𝑡 ∗ 𝜀𝑖
(17)
Where fertt is actual soil fertility and  is a constant depletion rate
(0-1). The soil fertility dynamic in agricultural crops is then:
15
𝑓𝑒𝑟𝑡𝑡+1 = min⁡(𝑓𝑒𝑟𝑡𝑚𝑎𝑥 , max⁡(0, 𝑓𝑒𝑟𝑡𝑡 − 𝛿𝑑𝑒𝑝,𝑡 + 𝛿𝑟𝑒𝑐𝑜𝑣,𝑡 ))
(18)
The crop yield in equation 16 represents potential value whereas
actual yield is limited by labor harvesting productivity and availability:
𝑝𝑜𝑡
𝑦𝑖,𝑗 = min⁡(𝑦𝑖,𝑗 , 𝜏𝑖 ∗ 𝐿𝑖 )
(19)
Where yi,j is actual or attainable yield of livelihood option i in plot
j (ton) and the total from all plots is used in equation 1 (see above), i
is harvesting productivity for option i (ton person day-1), and Li is the
available labor for option i calculated in equation 9. Potential yield of
non-agricultural crop systems are estimated based on mean and
variation between plots specified for each ecological stage of forest
types and for each production level of agroforestry or plantation
types. The actual or attainable yield depends on labor harvesting
productivity and availability as in equation 19. The potential
harvesting zones for NTFP and timber/logging are plots of any forest
type but not protected forest. Those for agroforestry products are
agroforestry plots at any production stage and for foods are plots of
any agricultural crop.
Food consumption and storage
The actual attainable yield and food storage will be used to satisfy
domestic consumption. The latter is the product of total population
(H, person) and food requirement per capita (, ton per capita year-1)
calculated for each food type:
𝐹𝑐𝑜𝑛𝑠,𝑖 = 𝐻 ∗ ⁡ 𝜃𝑖
(20)
In case of shortage, the community will buy the necessary
product from the market and actual buying depends on current
financial capital:
16
𝐹𝑏𝑢𝑦,𝑖 = min⁡(𝐹𝑐𝑜𝑛𝑠,𝑖 − 𝐹𝑠𝑡𝑜𝑟𝑒,𝑖 ∗ (1 − 𝑓𝑙𝑜𝑠𝑠 ),
𝑀𝑐𝑎𝑝,𝑡
ℎ𝑖
) (21)
Where Fstore,i is current storage of food i (ton), Mcap,t is current
financial capital (Rp), hi is product price (Rp ton-1) and floss is food loss
fraction (0-1), e.g. due to pests. In case of surplus, a portion will be
kept in the storehouse and the rest will be sold to the market.
𝐹𝑠𝑒𝑙𝑙,𝑖 = (𝐹𝑠𝑡𝑜𝑟𝑒,𝑖 ∗ (1 − 𝑓𝑙𝑜𝑠𝑠 ) − 𝐹𝑐𝑜𝑛𝑠,𝑖 ) ∗ 𝑓𝑠𝑒𝑙𝑙,𝑖
(22)
Where fsell is a fraction of surplus food will be sold to the market.
Food security index (0-1) is measured as follows:
𝑋𝑓𝑜𝑜𝑑,𝑖 = min⁡(1,1 −
𝐹𝑐𝑜𝑛𝑠,𝑖 −(𝐹𝑠𝑡𝑜𝑟𝑒,𝑖 +𝐹𝑏𝑢𝑦,𝑖 −𝐹𝑠𝑒𝑙𝑙,𝑖 )
𝐹𝑐𝑜𝑛𝑠,𝑖
)
(24)
Income from selling foods will be used to cover total cost and the
rest will become part of financial capital after allocating for
secondary consumption and land expansion:
𝑀𝑐𝑎𝑝,𝑡+1 = (𝑀𝑐𝑎𝑝,𝑡 + 𝑀𝑖𝑛𝑐,𝑡 − 𝑀𝑏𝑢𝑦,𝑡 − 𝑐𝑡 ) ∗ (1 − 𝑓𝑠𝑒𝑐 ) ∗ (1 −
𝑓𝑒𝑥𝑝 )
(25)
Where Minc is income from selling foods to the market (Rp), Mbuy
is expense for buying foods in the market (Rp), c is total cost (Rp) as
the sum of non-labor and labor cost as explained before in equation
1 and 2, fsec is a fraction of income for secondary (i.e. non-food)
consumption (0-1), and fexp is a fraction of income allocated for
establishing new plots. All non-agricultural crop products will be sold
to the market and thus Minc is total income from selling both
agricultural and non-agricultural products. In case of a disaster, loss
in financial capital is calculated according to a certain fraction.
17
Appendix 1. List of input maps and variables for FALLOW simulation
A. Maps
No
Name
Type
Description
1
Land-cover
and legend
Scalar
2
Soil and
geology, and
legend
Scalar
3
Area
boundary
Scalar
To produce a land-cover
map for FALLOW, we need
to reclassify the original
land-cover types into those
specified in FALLOW
interface (menu Customize)
registered in a scenario file.
We need therefore a
source legend and specify
the scenario name
containing the referenced
land-cover names. We
need also to specify if the
land-cover types have been
logged or not by checking
the ‘logged’ check box. This
will produce a logging map
of Boolean type.
Soil and geological map and
their legend are used to
produce inherent soil
fertility map. One or two
assessors need to do a
scoring (usually 1-5, 1=
poor, 5=excellent) for the
soil and geological types.
An actual soil fertility map
is produced through setting
fractions from inherent soil
fertility.
Option ‘fill no data’ is
provided to set ‘no data’ in
the map into a unit value
Name of input
file
Initlc.xxx
Initsoil.xxx for
inherent soil
map and soil.xxx
for actual soil
map
Area.xxx
18
4
Forest reserve
boundary
Boolean
5
Elevation
m. asl
6
Distance to
road
m
7
Distance to
river
m
Option ‘no forest reserve?’
is provided for a region
without forest reserve
We need an elevation map
to produce a slope map (in
degree)
Option ‘No available
objects to measure
distance?’ should be
checked when there is no
referenced object from
which distances are
measured. In that case, the
map will contain values of
1e10 which means every
plot has a very far distance
to the referenced object.
For a dynamic simulation of
road, we need to prepare 4
maps of distance to road, at
maximum, for 4 different
time intervals during
simulation period.
Idem with no 6
8
Distance to
market
m
Idem with no 6
9
Distance to
settlement
m
Idem with no 6
10
Distance to
processing
industries
m
Idem with no 6
Reserve.xxx
Slope.xxx
Droada.xxx,
droadb.xxx,
droadc.xxx,
droadd.xxx for 4
different
simulation
periods
Drivera.xxx,
driverb.xxx,
driverc.xxx,
driverd.xxx,
Dmarta.xxx,
dmartb.xxx,
dmartc.xxx,
dmartd.xxx,
Dseta.xxx,
dsetb.xxx,
dsetc.xxx,
dsetd.xxx,
Dindaf1a.xxx, …,
dindaf5d.xxx,
dindfd1a, …,
dindfd5d.xxx,
Dindloga.xxx, …,
19
11
Suitability
Scalar
12
Subcatchment
boundary
Disaster
Scalar
(integer)
13
Boolean
We should prepare a
suitability map for each of
livelihood option. For this,
the spatial team can
directly produce those
suitability maps (0=no
suitable, 1=suitable) for
each of the livelihood
options, or we can use a
general suitability map
containing information of
available land-cover types
and we decide whether
each of the land-cover
types is suitable or not for a
certain livelihood option by
selecting the check box
‘suitable?’
Map describing the subcatchment boundary
Map describing the area of
a disaster event where the
affected areas will be
directly converted to
pioneer forest
dindlogd.xxx,
Dindnta.xxx, …,
dindntd.xxx
Staf1.xxx, …,
staf5.xxx,
stfood1.xxx, …,
stfood4.xxx
Subcatch.xxx
Disaster.xxx
B. Variable inputs (biophysics)
No
Name
Unit
Description
Name of input
file
Soil
1
Depletion rate
% unit
yield-1
Years
Soil fertility depletion rate to
produce a unit yield
Period needed to achieve
half of inherent soil fertility
Soil.par
2
Half time for
recovery
Soil.par
20
Environmental service
3
Aboveground
biomass
Ton ha-
4
%
Floor biomass
5
1
Probability of
fire escape
Succession and growth
6
Time bound
(0-1)
7
Year
Initial age
Year
Yield
8
Yield
agroforestry
Ton ha-
9
Ton ha-
Yield NTFP
1
1
10
Logging
m3 ha-1
11
Agriculture
Ton ha1
Average aboveground
biomass for a certain landcover type
Fraction from aboveground
biomass
A probability fire spreads
from adjacent plots
Lces.par
Lower time bound for a
certain land-cover type
Initial age of land-cover
types
Lctime.par
Yield in agroforestry plots at
4 successive stages (pioneer,
early production, late
production, post
production)
NTFP Yield at 4 successive
stages of forest (pioneer,
young secondary, old
secondary, and primary
forest)
Logging yield at 4 successive
stages of forest
Agriculture yield for each
type of agriculture
Yield.par
Lces.par
Lces.par
Initlcage.par
Yield.par
Yield.par
Yield.par
C. Variable inputs (Social economic)
No
Name
Harvesting
1
Harvesting
productivity
Unit
Description
Name of input
file
Unit
yield
person
day-1
Harvesting productivity for
each livelihood option
Harvest.par
Establishment
21
Establishment
cost
3
Labor
requirement
Learning
4
Population
fraction
Rp ha-1
5
Alpha learning
(0-1)
6
Beta learning
(0-1)
7
Prioritization
degree
[]
2
External labor
8
External labor
involved
Demographic
Person
day ha-1
100%
Person
day
Cost required to open the
land minus the labor cost
Labor needed to clear land
for each livelihood option
Estcost.par
Population fraction for two
different types of farmers
(agent 1 and 2). E.g. agent
1 = conservative farmers,
agent 2 = modern farmers
Adjustment rate based on
current year experience for
subsequent year land-uses.
0 = farmers ignore current
year experience, 1=
farmers fully use current
year experience
Adjustment rate related to
suggestions from others for
subsequent year land-uses.
0 = farmers ignore
suggestions from others, 1=
farmers fully use
suggestions from others
Describing preference in
allocating financial or labor
resource to available
livelihood options for
subsequent year land-uses.
0=available resources will
be allocated uniformly
among livelihood options.
>1 = available resource will
be mostly allocated to the
most profitable livelihood
option
Agent.par
Labor from outside of
simulated area
Extlab.par
Estlab.par
Agent.par
Agent.par
Agent.par
22
9
10
11
12
Initial human
population
Annual
population
growth rate
Labor force
fraction
Annual working
days
13
Initial financial
capital
14 Secondary
consumption
fraction
15 In case of
disaster, human
population
decrease
16 In case of
disaster,
financial capital
decrease
17 In case of
disaster,
working day
decrease
Initial knowledge
18 Expected payoff
to labor
19
Expected payoff
to land
Fire use
20 Probability of
using fire for
land clearing
Yield storing
21 Consumption
Person
%
100%
day
person-1
year-1
Rp
100%
Initial human population in
the simulated area
Of population in the
simulated area
Demo.par
Fraction of labor from
population
Of labor in the simulated
area
Demo.par
Demo.par
Demo.par
Demo.par
Fraction of saved money
used for secondary
consumption
Decrease in human
population due to a
disaster
Demo.par
%
Decrease in financial capital
due to a disaster
Socdis.par
%
Decrease in working day
due to a disaster
Socdis.par
Rp
person
day-1
Rp ha-1
For each farmer type
Intknow.par
For each farmer type
Intknow.par
(0-1)
Slash and burn for clearing
land?
Fire.par
Unit per
capita
Collected yield used for
consumption
Store.par
%
Socdis.par
23
22
Probability to
sell
(0-1)
23
Loss fraction
100%
Cultural value
24 Cultural value
Extension
25 Extension
availability?
26 Extension
credibility
27 Exposure
fraction
Extension suggestion
28 Payoff to labor
29 Payoff to land
Subsidy
30 For
establishment
31 For maintenance
Market
32 Price
Non-labor cost
33 Non-labor cost
NTFP and
logging
34
Non-labor cost
agriculture
Probability that the
unconsumed yield will be
sold to the market
Fraction of collected will be
loss, e.g. due to pest
problem
Store.par
(0-1)
Non-economic
consideration for
subsequent year livelihood
options
Culture.par
0 or 1
0=no extension available,
1= extension available
Credibility assessed by
farmers to the extension
Farmer exposure for an
extension
Extense.par
Rp
person
day-1
Rp ha-1
Suggested by the extension
Extknow.par
Suggested by the extension
Extknow.par
Rp
Subsidy for plot
establishment
Subsidy for plot
maintenance
Subsidy.par
Rp unit-1
Price for each harvested
products
Price.par
Rp ha-1
Non-labor cost for each of
4 successive stages of
forest types (pioneer,
young secondary, old
secondary, and primary
forest)
Non-labor cost for each of
agricultural type
Cost.par
(0-1)
100%
Rp
Rp ha-1
Store.par
Extense.par
Extense.par
Subsidy.par
Cost.par
24
Rp ha-1
Non-labor cost for each of
4 successive stages
(pioneer, early production,
late production, post
production) of agroforestry
plots
Cost.par
(0-1)
Fractions describing the
importance of spatial
aspects considered in land
expansion (i.e. soil fertility,
plot utility, suitability of
land, transportation cost,
maintenance cost, land
clearing cost due to slope,
land clearing cost due to
floor biomass)
Spatial.par
Time series
37 Price
Rp unit-1
38
Extension
availability
0 or 1
39
Subsidy
0 or 1
Dynamic price (4 time
intervals during simulation
period). Option ‘on’ or ‘off’
is available for input time
series used
Dynamic extension
availability (4 time intervals
during simulation period).
Option ‘on’ or ‘off’ is
available for input time
series used
Dynamic subsidy
availability (4 time intervals
during simulation period).
Option ‘on’ or ‘off’ is
available for input time
series used
35
Non-labor cost
agroforestry
Expansion determinant
36 Expansion
determinant
25
Appendix 2. Minimum inputs for FALLOW simulation
A. Maps
No
1
2
Name
Land-cover
Soil and geology
Type
Scalar
Scalar
3
4
Scalar
Boolean
5
6
Area boundary
Forest reserve
boundary
Elevation
Distance to road
7
Distance to river
m
8
Distance to market
m
9
Distance to settlement
m
10
Distance to processing
industry
Suitability
m
11
m. asl
m
Scalar
Description
Map of land-cover type
This to produce map of soil fertility
index
Simulation area
If any protected forest
To produce a slope map
Distance of each pixel to the closest
road
Distance of each pixel to the closest
river
Distance of each pixel to the closest
market
Distance of each pixel to the closest
settlement
Distance of each pixel to the closest
processing industry
Suitability map for each simulated
livelihood options. (0=pixel not
suitable, 1=suitable)
B. Economic Input variables
Establishment cost
Labor requirement for
establishment
Initial financial capital
USD ha-1
Person day
ha-1
USD
100%
5
Secondary
consumption fraction
Consumption
6
Return to labor
1
2
3
4
Unit per
capita
USD
Financial capital in the simulated
landscape
Fraction of saved money used for
secondary consumption
Collected yield used for consumption
26
7
8
9
10
11
Return to land
Subsidy for
establishment
Subsidy for
maintenance
Price
Non-labor cost
person
day-1
USD ha-1
USD
USD
USD ton-1
USD ha-1
C. Biophysical and demographic variables
No
1
Name
Aboveground biomass
Unit
Ton ha-1
2
Age range of growth
stage
Year
3
4
Yield
Harvesting
productivity
5
Initial human
population
Annual population
growth rate
Labor force fraction
Annual working days
Ton ha-1
Ton
person
day-1
Person
6
7
8
Description
Average aboveground biomass for
each land-cover type
The simulated livelihood options
consist of pioneer, early, mature, and
post production. We need to specify
the age range for these stages
Yield for each stage
Harvesting productivity for each
livelihood option
%
100%
Person
day-1 year-
Fraction of labor from population
Of labor in the simulated area
1
27
Appendix 3. Running FALLOW model with Nutshell
There are some advantages of using Nutshell to run the FALLOW
model:
• More options for model settings
• More options for displaying results
• Possibility to change the model code
• Runs under different operating system versions, which may not
be compatible with the FALLOW GUI (e.g. Windows Vista)
• Detailed report when the model crashes
The followings are steps in running FALLOW with Nutshell:
1. The interface of Nutshell is as shown in Figure 7. It opens by
executing the Nutshell.exe file. For the first time use, it usually
asks the location of PC Raster/apps folder. So please set this path
in the dialog box that appears when opening Nutshell for the first
time
2. Next step is to select the model folder. If the FALLOW model is
located in C:/ for example, then we need to select this folder and
click the button for setting the working directory (no 4 in Figure
7)
3. To run the model we need to open the Fallow8AF.mod file.
Please find this in the model folder and open it by clicking the
button ‘Edit a model’ above the explorer window. The contents
of the file appear like shown in the right window of Nutshell in
Figure 7. The file contains the code of FALLOW model written in
PC Raster language
4. The second line in the code specifies the length of simulation in
year. Please modify accordingly and save the file before running
5. To run the model, please select the green arrow head (no 6 in
Figure 7). If the model runs well, it will appear ‘Executing time
step 1’ in the above left window and all outputs will be stored in
28
the same folder where we store all inputs (for example C:/Fallow
like above)
6. The arrow in the above left window indicates the place to write
PC Raster commands (i.e. to do map operations). Please read the
PC Raster manual for descriptions of each PC Raster command
and example. Appendix 4 below also describes some important
commands for map operation
Figure 7 Nutshell interface as a user friendly way to run and edit the FALLOW model
29
7. All files with *.xxx extension are input maps for FALLOW model.
The extension .xxx indicates that it can be modified by users. But
modification should be made consistently too in the FALLOW
code (i.e. Fallow8AF.mod). All input data are in .par format.
Please see again Appendix 1 for the list of input maps and data
required by the FALLOW model. Please see Appendix 4 for
instruction of how prepare the input maps and data
8. All output variables are with *.out extension. For example
lcarea.out-af1_pion is for time series output of area of
agroforestry system type 1 at pioneer stage. The output
landcover maps are with the landcovr.* name
9. To extract important outputs, we provide an Excel file
‘output.xlsm’ that can display output files from the model folder.
Please open this file, set the model folder and then activate the
import button
30
Appendix 4. Important notes in running FALLOW
A. Preparing input data
1. Check if all Ascii maps have the same attributes
2. If yes, please prepare the clone map (see below how to do
this)
3. Then convert all maps to .xxx format (see below how to do
this). Please see the FALLOW code so we know what input
maps that we need to prepare
4. After preparing maps, please see the FALLOW code that lists
each *.par file to input socio economic and biophysic
parameter values
B. Map operations in PCRaster
1. Create a clone map
Please find tmp.bat from the FALLOW directory that contains
the following commands for creating a clone map:
mapattr -s -R 1036 -C 840 -S --small -P yt2b -x
544247.37365491 -y 2411507.6846066 -l 100 clonemap2.tmp
asc2map -a -m -9999 --clone clonemap2.tmp
BK_border_1ha.asc area.rmp
display area.rmp
Please adjust the map attributes accordingly and type tmp in
the Nutshell dialog box.
31
2. Change from Ascii maps to FALLOW maps
For example: asc2map --clone area.map -a -S lc_backan.asc
lc_backan.xxx
3. Change from FALLOW map to Ascii maps
For example: map2asc -a -m 9999 backan.xxx backan.asc
4. Conditionality
For example: pcrcalc test2.map = if(test1.map eq 1,2,
if(test1.map eq 2 or test1.map eq 3 or test1.map eq 4,3,
if(test1.map eq 6, 4, if(test1.map eq 7 or test1.map eq
8,5,0))))
5. Change a missing value to a value
For example: pcrcalc test.xxx=cover(test1.xxx, 0)
6. Running the FALLOW model
There are two ways:
 Open the FALLOW code and then run the model by
clicking the ‘run’ button
 With the PCRaster command in the Nutshell dialog
box:
pcrcalc -f fallow08.mod
7. Error that comes out if format of the input maps are not
similar
32
For example:
C:\FALLOW>pcrcalc -f fallow08.mod
pcrcalc version: May 31 2001 (win32)
fallow08.mod:530:9: ERROR: droadb(binding=droadb.xxx):
location attributes of 'area.xxx' and 'droadb.xxx' are different
The solution is that to make a new clone map that matches
the other input maps, or in case that the attributes of the
input maps are different each other, then try to resample the
input maps according to the clone:
C:\FALLOW>resample --clone area.xxx droada.xxx test.xxx
resample version: May 7 2001 (win32)
C:\FALLOW>copy test.xxx droada.xxx
Overwrite droada.xxx? (Yes/No/All): y
1 file(s) copied.
8. To clean the outputs of FALLOW running the type clean in the
Nutshell dialog box
9. Display a map with its legend
display -p lc.pal BK_landcover.map
10. Display maps dynamically after long time simulation
For example to display the map along the year after 20 year
simulation with the FALLOW model: display -p lc.pal
landcovr.001+20
11. To make a slope map from the elevation map
pcrcalc slope.xxx=slope(elevation.xxx)
33
12. To calculate distance from each pixel to settlement
dset=cover(spread(lc eq lcid[set],0,1),1e11)*area;
34
References
Lusiana, B., van Noordwijk, M., Cadisch, G. 2012. Land sparing or
sharing? Exploring livestock fodder options in combination with
land use zoning and consequences for livelihoods and net carbon
stocks using the FALLOW model. Agriculture, Ecosystems and
Environment 159: 145– 160
Lusiana, B., van Noordwijk, M., Johana, F., Galudra, G., Suyanto,
Cadisch, G. 2013. Implication of uncertainty and scale in carbon
emission estimates on locally appropriate designs to reduce
emissions from deforestation and degradation (REDD+). Mitig
Adapt Strateg Glob Change. doi:10.1007/s11027-013-9501-z
Mulia, R., Widayati, A., Suyanto, Agung, P., Zulkarnain, M.T. 2013.
Low carbon emission development strategies for Jambi,
Indonesia: simulation and trade-off analysis using the FALLOW
model. Mitig Adapt Strateg Glob Change. doi: 10.1007/s11027013-9485-8
van Noordwijk, M., 2002. Scaling trade-offs between crop
productivity, carbon stocks and biodiversity in shifting cultivation
landscape mosaics: the FALLOW model. Ecol. Model. 149, 113–
126.
van Noordwijk, M., Suyamto, D.A., Lusiana, B., Ekadinata, A. and
Hairiah, K. 2008. Facilitating agroforestation of landscapes for
sustainable benefits: tradeoffs between carbon stocks and local
development benefits in Indonesia according to the FALLOW
model. Agriculture Ecosystems and Environment Special Issue on
Climate Change Research CGIAR.
Suyamto, D.A., van Noordwijk, M., Lusiana, B., Ekadinata, A. and
Khasanah, N. 2005. Prospects of adoption of tree-based systems
in a rural landscape and its likely impacts on carbon stocks and
farmers’ welfare: the FALLOW Model Application in Muara
Sungkai, Lampung, Sumatra, in a “Clean Development
Mechanism’ context. ICRAF Working Paper WP06034.
35
Suyamto, D. A., Mulia, R., van Noordwijk, M., Lusiana, B. 2009.
FALLOW 2.0. Manual and Software. World Agroforestry Centre,
Bogor, Indonesia
Tata, H.L., van Noordwijk, M., Ruysschaert, D., Mulia, R., Rahayu, S.,
Mulyoutami, E., Widayati, A., Ekadinata, A., Zen, R., Darsoyo, A.,
Oktaviani, R., and Dewi, S. 2013. Will REDD+ funding stop peat
swamp conversion to oil palm in orangutan habitat in Tripa
(Aceh, Sumatra, Indonesia)? Mitig Adapt Strateg Glob Change.
doi: 10.1007/s11027-013-9524-5
36