76 C. Line Carpentier et al. Agriculture, Ecosystems and Environment 82 2000 73–88
Fig. 2. Proportion of 1996 farms in each land use by age of farm settlement.
value of total output in 1994 was extracted from pas- ture and livestock activities milk and beef, while only
8 of farmers’ value of total output derived from the forest mainly Brazil nuts since timber extraction is
limited by administrative requirements making timber extraction practically impossible.
Against this backdrop of land-use systems and income-generating activities, what does it mean to
intensify? Intensification can mean 1 increasing the amounts of purchased inputs dedicated to a particular
activity indicated by a box in Fig. 1, 2 increas- ing the amount of labor dedicated to the activity,
3 increasing output per unit of land, 4 combining some of the above, 5 extending the useful life of a
Fig. 3. Settlement of farmers’ on-farm income sources value of total output.
particular land use, or 6 increasing the number of ‘arrows’ linking activities, thereby providing farmers
with more options. In this study and in the model used for the analysis,
intensification is viewed in an activity-specific way — products can be produced using different levels of
intensities. Higher levels of intensity usually but not always mean increased returns to land and labor, ad-
ditional purchased inputs, and increased sustainability in the sense that the land use ‘lasts longer’. Given
scarce labor and limited access to credit but relatively abundant land in the area, it is not clear a priori
whether these intensive systems will be adopted.
3. Analytical tool: a farm-level bioeconomic model
To understand the factors driving land use and deforestation at the meso-level, actors at the micro-
level making decisions that result in the observed meso-level land uses and the reasons behind their
decisions must be understood. Once those are under- stood, specific actions policymakers and technology
designers might take to alter farmers’ decisions can be identified.
Settlement farmers are only one of the three main land-use decision-makers in Acre. This study focuses
on settlement farmers, in part because of their some- times impoverished state which itself merits policy
action, but more importantly because of their critical current and future roles in deforestation. Though these
farmers are small on the Brazilian scale of operations, they are not small by international standards, averag-
ing 83 ha in the sample though the sample was limited to farms with less than 200 ha, some larger farms can
be found within the settlement projects. These farm- ers are not welfare poor; they generally have enough to
eat, especially once the establishment phase of farm- ing in the forest margins has passed. Annual land-use
decisions of these settlers, estimated to approach half a million in the Amazon, can have significant im-
pacts on total forest conversion. It is thus pivotal to understand these farmers’ reactions to combinations
of intensification types, policies, and institutional arrangements to predict their deforestation and eco-
nomic implications before they are made available to farmers.
C. Line Carpentier et al. Agriculture, Ecosystems and Environment 82 2000 73–88 77
3.1. FaleBEM — the bioeconomic model The biophysical and socioeconomic complexities
faced by settlers were integrated in a farm-level bioeconomic FaleBEM linear programming LP
model using survey data supplemented by techni- cal coefficients collected via farmers and experts.
FaleBEM is discretely dynamic in the sense that each year’s activities result in a new stock of financial and
natural resources that become the initial conditions for the next year. Nutrients, cash, livestock, pasture,
and fallow of different ages and technology levels, for instance, can be carried forward to the next season,
used, or sold in any season. The optimal solution is the one that maximizes the discounted value of the
family’s consumption stream over a 25-year time horizon by producing a combination of products for
home consumption and sale, selling labor off-farm, or extracting products from the forest subject to an
array of socioeconomic, policy, and biophysical con- straints. Although FaleBEM maximizes consumption
resulting from the intertemporal allocation of in- comes and savings, the term “income” is used in the
text because it is a more familiar concept than con- sumption. Socioeconomic constraints reflect prices
and input and output market imperfections in the settlement. For example, quotas are imposed on milk
sales to reflect the decreasing price schedule offered by milk processors, monthly labor available to hire
is limited to 15 man-days to reflect the scarcity of labor, wages vary monthly to reflect peaks in labor
demand, and expenses are limited to the seasonal amount of cash in hand to reflect the difficulty in
obtaining credit. Markets and policies are reflected in prices and the set of activities permitted in the
simulation. For example, the 50 rule mandating that no more than half of any farm be cleared for
agricultural purposes is not enforced in the simula- tions to indicate that the law has not been enforced
on settlement farms so far. FaleBEM includes all of the common crops and crop rotations intercropped
or grown in succession at various levels of intensi- fication found in the settlement. Crops include rice,
beans, maize, coffee, bananas, manioc, and grass and legume pasture. Technologies had to be packaged
because of data limitation, thus preventing piece- meal adoption of intensification technologies in the
simulations. Biophysical factors constrain activity choices over
time. More specifically, biophysical constraints re- quire that enough nutrients are available to cultivate,
or else yields decrease through a linearized nutrient response function. Also, agronomic constraints limit
possible land-use sequences. For instance, annual crops cannot be planted after pasture because tillage
is not used on settlement farms. Soil fertility is man- aged either by adding commercial fertilizer, changing
the product mix, letting land go fallow, or expanding into new areas deforesting.
Given the complexity of the settlers’ land-use decisions and the limited understanding about these
decisions at the beginning of this study, general equi- librium effects have not been included in the model
yet. The effect of ignoring general equilibrium output effects should be limited under current conditions
plans are to include these effects following Bouman et al., 1998. For instance, the major products being
selected by the model, beef, is traded internationally and regional demand is still not completely satisfied
by regional supply Faminow et al., 1996. Likewise, milk-processing capacity is substantially underutilized
Oliveira et al., 1999. Thus, the change in price re- sulting from increased supply may be relatively small.
However, if non-timber forest products, which face notoriously thin and seasonal markets, were an im-
portant activity selected by the model, price elasticity of supply would have to be used to reflect a decrease
in price as supply increases. Potential general equi- librium effects on the input side, especially labor and
wages, were addressed through sensitivity analyses. Two other limitations of the model are that no risks are
included and land cannot be rented, purchased or sold.
Basically what this farm-level model does is present the farmer with the complete set of land-use options
and intensity levels available that are changed to gen- erate different simulations and then performs several
‘reality checks’ that constrain farmer decisions, such as input requirements, input availability, reversibility
of land-use decisions, and profitability. The model se- lects, from all possible land-use paths over a 25-year
period, the one that maximizes the discounted sum of income. So, of the many possible arrows and boxes
appearing in Fig. 1, the model chooses the best collec- tion of both. The chosen land uses and intensities are
thus profitable compared to other activities and could be expected to be adopted by farmers.
78 C. Line Carpentier et al. Agriculture, Ecosystems and Environment 82 2000 73–88
3.2. Model data Three types of data were needed to build FaleBEM:
1 price, cost, and market data, 2 biophysical data yield, nutrient demands, nutrient accumulation in the
fallow, etc., and 3 data about farmers’ initial con- ditions. Price, cost, and market data were collected
from public sources and completed through interviews with farmers, transporters, and buyers. Biophysical
data were gathered through the ASB program Munoz Braz et al., 1998; Palm et al., 1996, focus groups, in-
terviews with experts in the area, and published data e.g., Valentim Ferreira, 1990.
Because the model is dynamic it must be initiated with initial conditions about the household and the
farm being modeled. Each farm in the sample de- scribed in Section 2.3 can be modeled by changing
the model’s initial conditions. However, the task was simplified by clustering the 1994 data based on char-
acteristics exogenous to farmers’ decisions, such as
Table 1 Farm and farm household initial conditions in 1994 for two farm types
Initial conditions year 0 of model simulations Well-situated farms
Distant farms Household assets and liquidity
Adult male laborers man-days 1.63
1.42 Other family laborers adult male equivalents
1.47 1.05
Pension Rmonth
a
120 Food storage capacity kg
2000 2000
Initial cash balance R 250
250 Initial forest reserve ha
43 67
Initial cleared land ha 17
22.5 Labor, markets, and credit
Labor transactions maximum number of man-daysmonth Hired
15 15
Sold 15
15 Milk quota maximum liters soldday
50 50
Agricultural credit R Transportation
Transport time daysround trip to market Ox, dry season
1.5 2.88
Ox, rainy season 2.0
3.83 Truck, dry season
0.63 1.0
Truck, rainy season 0.78
1.0 Transport cost round trip to market
Truck R 91
104 N
b
25 26
a
All prices are reported in terms of 1996 Brazilian reais.
b
The sample size is smaller because it includes Acre farms only and only those farms falling within the two clusters are presented in the table.
soil type, distance to market, and age of the farm. Sev- eral clusters of farmers with similar patterns emerged,
each of which can be thought to represent a farm type. The set of initial conditions used in this study
represents well-situated farms in terms of market access with predominant medium-quality soil types
those with some fertility problems andor mild slope or rockiness. Well-situated farmers were emphasized
here because they will become more prevalent in the future as technology spreads, infrastructure improves,
and unsuccessful farmers give up and are replaced by more successful ones. The initial conditions for that
farm type and distant farms are presented in Table 1. Initial land-use conditions are also captured at year
“zero” along the horizontal axis of Figs. 4–7, which depicts that in 1994 a majority of the lot was forested
43 ha and the lot size was 60 ha.
Other parameters collected during the fieldwork are used in the model after they were calibrated with
groups of experts, such as land-use inputs including
C. Line Carpentier et al. Agriculture, Ecosystems and Environment 82 2000 73–88 79
Fig. 4. Land uses when no technological intensification is available.
Fig. 5. Land uses when intensification is available for non-livestock activities on cleared land.
Fig. 6. Land uses when intensification is available for all activities on cleared land.
80 C. Line Carpentier et al. Agriculture, Ecosystems and Environment 82 2000 73–88
Fig. 7. Land uses when intensification is available for all activities on cleared and forested land.
monthly labor requirements and output levels see Carpentier et al., 2000a, for details. The model was
validated using the panel data collected for 1994 and 1996 and by modeling the various farms from the ini-
tial settlement when the lot was completely forested to the present and by determining if the modeled land
uses and land-use paths resemble those currently ob- served in the field. Sensitivity analyses — how stable
or robust is the solution to factors varied one at a time — were conducted on six factors: soil quality, labor,
prices, discount rate, distance to market, and access to market. Labor availability, distance to market, and
access to market especially to milk processors were found to affect deforestation rate, but none of the other
factors did. Prices and market access greatly affect uses of the cleared land and income, as expected. All
factors except discount rates affect income. Discount rate had little effects on results because investments
with returns in later years, such as perennial crops, were not selected under 1994 conditions. Though the
magnitude of the effect varied by farm type, the sign of the changes was the same for all farm types.
This validated baseline reflecting 1994 policy, socioeconomic, technological, and biophysical con-
ditions can be used to predict what would happen if one or a combination of these conditions were al-
tered. Before presenting these simulations, however, more information is presented on the intensification
alternatives present in the model.
3.3. Modeled intensified systems Currently, intensification strategies are mainly avail-
able for cleared land: for annual crops, perennial crops, and pasture. However, a pilot study conducted by EM-
BRAPA Brazilian Agriculture Research Corporation provides the information necessary to model intensifi-
cation in the forested area, represented by low-impact forest management.
3.3.1. Cleared land Our research effort and collaboration with EM-
BRAPA uncovered sets of intensification strategies for annual and perennial crop production, as well
as pasture for livestock production. For example, Table 2 presents the extensive low-production pas-
ture system and an intensified system recommended by EMBRAPA. Intensified pastures are seeded using
improved purchased seeds and demand more labor to build more fences, thus improving pastures’ carrying
capacity. These practices increase the yearly carrying capacity and increase the useful life of the pasture by
slowing the decay in carrying capacity. Increasing the intensity of the livestock operations in this context
means increasing labor to improve management and capital inputs to provide more veterinary and supple-
mental feeds, yielding higher calving rates, higher milk production and duration of the lactation period,
and lower mortality rates Table 3. Similar technol- ogy tables are available for annuals, perennials, and
deforestation.
3.3.2. Forested land Preliminary results of a low-impact forest manage-
ment pilot study being conducted by EMBRAPAAcre are incorporated into the model. Under current leg-
islation it is practically impossible to sell extracted timber from on-farm reserves, making this standing
C. Line Carpentier et al. Agriculture, Ecosystems and Environment 82 2000 73–88 81
Table 2 Technical coefficients for settlement farms
1
pasture production systems in Acre, by pasture type and level of technology Technical coefficients
Grass Grasslegume intensive
Extensive Intensive
Pasture establishment and management Seeds kgha
Brizantao 15
15 15
Kudzu 1
Labor man-daysyear Seeding year 1
3 3
3 Weeding year 1
2 3
3 Weeding years 2–4
2 3
3 Weeding years 5–11
2 3
3 Fencing
Length m of fenceha of pasture 63
106 106
Oxen time man-dayskm of fence 4
4 4
Own chainsaw man-dayskm of fence 4.5
1 1
Labor man-dayskm of fence 59
56 56
Total costs Rkm of fence
a
302 347
347 Pasture productivity
Carrying capacity animal unitsha, rainy season Year 2–3
1 1
1.5 Year 4
1 1
1.5 Year 5
0.88 0.99
1.5 Year 6
0.79 0.97
1.5 Year 8
0.49 0.9
1.5 Year 9
0.39 0.85
1.5 Year 10
0.29 0.8
1.5 Year 11
0.3 0.85
1.48 Year 15
0.65 1.4
Year 20 0.9
a
All values are in 1996 Brazilian reais, labeled R; one R was roughly equivalent to one US.
forest of little value to farmers. Allowing farmers to extract timber in a sustainable way would increase
the value of the standing forest and remove incentives farmers have otherwise as will be demonstrated in
the following section to clear the forest completely despite the 50 rule. The rule is not effective because
the Brazilian authorities have not had the administra- tive capacity to enforce this law on the half-million
settlement farms found in the Amazon.
EMBRAPA obtained permission from the environ- mental ministry IBAMA to train 10 selected farmers
to practice low-impact forest management on their on-farm forest reserves. Together, the farmers could
afford the initial investments necessary to start extract- ing timber while minimizing environmental impacts.
Each farmer had on an average 40 ha of legal reserve, thus the total extraction area was 400 ha. The common
area of legal forest was divided into 10 subplots of 40 ha harvested on a rotational basis, 40 ha per year or
4 ha per farmer, leaving each harvested area 10 years to regrow. The extraction was done at a low-intensity
level, 10 m
3
ha
− 1
per year or approximately 1.5 trees ha
− 1
, compared to 30–40 m
3
traditionally ex- tracted by loggers in the area and only of trees larger
than 50 cm in diameter. Low-impact forest manage- ment also minimizes damages to the forest through
improved felling techniques and by dragging the wood with oxen instead of tractors de Araujo Borges and de
Oliveira, 1996; de Oliveira Neves et al., 1996. This technology is modeled by allowing farmers to extract
1 m
3
ha
− 1
each year farmers can extract 10 m
3
from 4 ha one-tenth of 40 ha, translating into 1 m
3
ha
− 1
. Labor requirements per cubic meter are 2.62 man-days
women and children cannot perform this task; 1.41
82 C. Line Carpentier et al. Agriculture, Ecosystems and Environment 82 2000 73–88
Table 3 Technological coefficients for settlement farms cattle production systems in Acre, by level of technology
Technical coefficients Extensive
Intensive Herd dynamics
Calving rate 50
67 Mortality rate
1 year 10
6 2 years
5 3
2 years 3
2 Cullingdiscard rate
Cows 10
Bull 6
12 Herd inputs
Feed supplements elephant grass, forage kganimal 20
Salt kganimalyear 10
Mineral salt kganimalyear 18.25
Animal health Aftosa vaccinationsanimalyear
2 2
Brucellosis vaccinationsfemale calfyear 1
Rabies vaccinationsanimalyear 1
Carrapaciticida ml of butoxanimalyear; to control ticks 5
10 Worm control mlanimalyear
10 25
Antibiotics Mata bicheira cubic centimeteranimalyear; to control internal parasites
0.03 Terramicina mlyear to
1 2
the herd 0.06
0.13 Labor for herd management
Milking man-dayslactating cowmonth 0.9
1.5 Other livestock man-daysanimal unitmonth
0.3 0.6
Herd production Meat production
Calf weight kg 60
75 Weight of fattened steer kg
210 225
Milk production Milk production dry lday
2.5 4.5
Milk production wet lday 3
6 Lactation period daysyear
180 240
man-days requires a chain saw and 0.2 man-days re- quires an ox. The remaining is just labor. Returns per
cubic meter or per hectare, excluding labor and equip- ment costs but including transportation costs, is R95.
4. Simulations and results