Environmental Modelling & Software 25 (2010) 1322e1333

Contents lists available at ScienceDirect

Environmental Modelling & Software

journal homepage: www.elsevier.com/locate/envsoft

Designing and implementing a Role-Playing Game: A tool to explain factors,

decision making and landscape transformation
Manuela Vieira Pak a, *, Daniel Castillo Brieva b
a
CIRAD-UPR GREEN, Montpellier, France and PhD student at AgroParisTech, Paris, France
b
School of Environmental and Rural Studies, Institutions and Rural Development Research Group, Javeriana University, Bogotá Colombia

a r t i c l e i n f o a b s t r a c t

Article history: In this paper we describe a research process on contextual driving factors and decision-making processes
Received 4 February 2008 used by local actors for land use change in a zone of the Colombian Amazonian frontier. We integrated
Received in revised form landscape multi-temporal analysis, Role-Playing Games (RPG), interviews based on flow diagrams and an
15 March 2010
historical study of landscape dynamics for the construction of our methodological approach. Findings of
Accepted 16 March 2010
Available online 31 May 2010
the study include individual detailed decision-making insights at the farm level that shed light on the
mechanisms that boost the advance of the agricultural frontier into the Amazonian forest. We illustrate
how individual decisions are related with the general landscape dynamics. A formalization of results was
Keywords:
Role-Playing Game
carried out in UML (Unified Modeling Language) for the future construction of a Multi Agent System
Agent based modeling (MAS) model, the implementation of which will be useful for land use planning, discussions among local
Colombian Amazonian frontier and regional actors and scenario building. The RPG constitutes a device that could “talk” by itself, in the
Participatory tools name of local actors. Facts that hardly would be communicated in an interview emerge implicitly and
UML explicitly through the exercise. The RPG is a device that we call a “dense methodological tool”, in the
Multi-temporal analysis sense that it is a designed object that synthesizes a complex system. This is central to territorial planning
Land use change because RPG and derived MAS models talk to actors and researchers in the same language that human
Landscape transformations
memory and projection mental capabilities function. These objects condense time and space and help
make problems clear, and they assist in the finding of solutions and exploration of possible scenarios.
Ó 2010 Published by Elsevier Ltd.

1. Introduction the relationships among transformation dynamics, factors, and

Throughout Colombian environmental history, human activities
have affected the components, structure and functioning of natural
ecosystems. Economic and social forces have been the drivers of
these transformations, which interact in a highly complex manner
(Palacios, 2001). In Amazonian colonization zones, transformations
have produced the establishment of pastures for cattle raising,
replacing and affecting the tropical forest.
Landscape dynamics studies, especially multi-temporal studies,
have been utilized to observe and to quantify land cover changes
and the tendencies of a specific research problem. However, in
Colombian contexts, the transformation factors of ecosystems, the
behavior of social actors regarding natural resource management
and the possible scenarios derived from specific decision-making
processes have not been studied in depth. There is a need to develop
and implement alternative methodologies that allow the study of
* Corresponding author.
E-mail addresses: manuela.vieira.pak@gmail.com (M. Vieira Pak), d.castillo@
javeriana.edu.co (D. Castillo Brieva).
decisions, as well as the underlying processes that lead to change.
The objective of this study, carried out in 2004e2005, was to
identify: 1) the factors that determine the decisions of a group of
colonos1 in a colonized area of the Department of Guaviare
(Colombian Amazon) during the establishment of pastures.2 These
factors refer to the identification of the contextual conditioning
elements that contribute to the process of land cover trans-
formation. These are the variables that affect the land use intensity,
resource access, and environmental, economic and cultural changes
that induce local actors’ decisions; and 2) the decisions that
generate a change in land cover from tropical forest to pasture.
Decisions refer to the actors’ decision-making processes and
includes decision rules and land use activities which in turn
generate land cover change.
1
In this study the term colono refers to peasants from other parts of Colombia
that migrate to Amazonian border zones.
2
We use the terms ‘pasture’ to specific land cover where grass vegetation is the
dominant form of plant life and ‘grassland’ to the specific land use where herbage is
suitable for grazing by livestock.

1364-8152/$ e see front matter Ó 2010 Published by Elsevier Ltd.

doi:10.1016/j.envsoft.2010.03.015
 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
The study of land cover change was carried out through a multi-
temporal analysis focused on forest and pasture change, based on
the guidelines of landscape ecology and the use of GIS and satellite
imagery interpretation from the years 1988, 1994 and 2001. With the
aim of understanding the links between spatial dynamics and social
interactions, we carried out interviews based on flow diagrams and
a role-playing game (RPG). Information generated by these tools fed
the conceptualization of a UML (Unified Modeling Language) model
used to begin a construction of a multi-agent system to explore the
land cover change problem in the zone. In this way we contributed
to building a tool that could be used by government agencies, such as
SINCHI3, to formalize information that they have acquired, and to
better understand the expansion dynamics of the agricultural
frontier. The tools utilized allowed an understanding of territorial
occupation, land uses types, actors’ decisions and their interrela-
tionships, as well as the production of information for territorial
planning and natural resource management.
2. Case study
During the last 30 years, a colonization zone has formed with cattle raising as
one of the dominant productive systems in the north Colombian Amazonian
region (Murcia, 2003). One of the more visible consequences of colonization in
tropical forests is the logging activity used to clear areas for the introduction of
cattle raising systems. The research problems addressed in the study zone include:
1) the loss of natural forest in order to add value to the land with pastures for
cattle raising and agricultural production (SINCHI, 1998); and 2) the tendency to
increase the area of pasture which is an inadequate productive system according
to the eco-biological conditions of the Amazonian environment. The last point has
been the perspective of SINCHI on the situation. SINCHI notes that cattle raising
production is a non-viable productive system in environmental, social and
economic dimensions. This lack of feasibility is partially due to the limitations in
market access, the deficiencies in communication and transport, and the poor soil
condition (Fajardo, 2002).
The Guaviare colonization area is located in the northern tip of the Colombian
Amazonian region covering approximately one million hectares (SINCHI, 1999).
The study zone extends over 48.719 ha. Guaviare region is a transition zone between
the savannas from the Orinoco to the Amazonian tropical rain forest ecosystems. The
region is characterized by a transition landscape of savannas, forests, grasslands,
coca crops and subsistence agriculture (see Fig. 1) (Ordóñez et al., 1989).
Colonos in the Guaviare colonization zone have settled in the Amazonian plains.
This territory is characterized by three different degrees of human intervention,
which are determined by the predominance of cultivated pastures (SINCHI, 1998): 1)
high human intervention zone (above 70%), 2) moderate intervention zone (between
40 and 70%), and 3) low intervention zone (below 40%). The peasant household is
a combination between productive and consumption units. It does not function as an
enterprise, but as a family unit which depends on its relationships with the socio-
cultural and natural environments (Forero et al., 2002). Although the colonos have
peasant origins, when they arrive in the colonization zone they integrate new attri-
butes into their culture. In new geographical spaces and environments, their ratio-
nality, in addition to maintaining the family unit, tends to maximize their utility for
self benefit (SINCHI, 1998). Therefore, a society exists in this zone with peasants-
colonos who have an atypical peasant rationality, one that is mixed (SINCHI, 1998).
We identified three types of actors, without taking into account indigenous groups:
1) The pioneer-colono, of peasant origin who generally comes from the
Colombian Andean zone. These people settle in the unconsolidated zone, and they
do not have land property titles even if they own the land by de facto occupation.
Social and productive actions are based on family relations (Salazar, 2002). 2) The
peasant-colono has a rationality of household maintenance where the basic unit is
the family and the goal is the territorial consolidation for familial reproduction. In
addition, this type of actor seeks to invest capital, buying land and increasing land
rent through the establishment of pastures. The peasant-colono can be considered as
a semi-entrepreneurial unit (Salazar, 2002). According to Salazar and Molano (in
Salazar, 2002) the unique way to avoid the peasant degradation is through coca
crops (although these are illegal) and cattle raising. 3) The latifundista4, according
3
Amazonian Institute of Scientific Research (Instituto Amazónico de Inves-
tigaciones Científicas).
4
The English translation of latifundista would be landlord, but the concept in the
region does not have to do with the fact that a person lets land to a tenant. In the
region, these people arrived as pioneer-colonos, then they become peasant-colonos
and accumulated a considerable quantity of land compared with the average
colonos in the zone. Locally, these people are called latifundista.
1323
Salazar (2002), has two possible origins. This actor could be either an enriched
peasant or a person who, traditionally, accumulates large land extensions. The large
state owner invests in establishing pastures for cattle raising. This actor has a high
capacity to contract a labor force.
Colonization zones in Colombia are affected by the internal armed conflict and
the Guaviare region is not an exception. The region has experienced periods of
violence and territorial domination by different armed groups. This characteristic
makes field work in the zone difficult and has consequences in the decision-making
processes of local actors. Regarding land use, a critical issue is the practice of growing
illegal products such as coca crops, which provide an important economic incentive
for colonos and armed groups. This form of land use is an important component of
landscape transformation, as will be demonstrated by the results of this study.
Methodologically, this situation posed a challenge because it was not possible to visit
the peasants-colonos on their properties, or to stay for a long period in the zone.
Consequently, the RPG tool provided an alternative to represent the conditions of the
region and the land cover changes at household level, in a safer environment such as
in the administrative center of the region, San Jose del Guaviare village.
3. Methodological framework
3.1. The integral approach
In order to meet the objectives of the study a four phase
methodology was designed. Fig. 2 portraits the methodological
framework.
During phase I we carried out the identification of: 1) land
covers, 2) actors, 3) decisions, 4) factors affecting the process of
establishment of pastures and 5) information needed for a prelim-
inary conceptual multi-agent system (MAS) model. The literature
review allowed the building of an historical profile of land cover
evolution in the colonization zone. This historical profile proved
useful in identifying the factors that influenced landscape trans-
formation, to define actors and their possible decisions regarding
the establishment of grasslands. In order to identify the land covers
we used Landsat satellite images5 supplied by SINCHI. We produced
a map of land cover units after cartographic processing. Finally,
based on collected information and semi-structured interviews, we
designed the tools that would used in the field. The principal
objective of RPG was to identify decisions and factors that influence
land cover transformation.
During phase II field work was carried out in order to: 1) verify the
types of land covers identified in the satellite image, 2) to collect
information for the RPG and, 3) to carry out semi-structured inter-
views to construct, with field informants, activity diagrams about the
activities that describe the land use transformation process from
forest to pasture. During phase III a land cover multi-temporal anal-
ysis was carried out with the purpose of quantifying changes in land
covers in two periods. Fig. 3 shows the results of the multi-temporal
analysis. The historical profile generated information about relevant
factors that have conditioned actors’ decisions identified in the
literature review. RPG, interviews and direct observation were
utilized to construct UML diagrams of most important activities. This
information was the input for the UML formalization of decision rules
of peasants-colonos, and the construction of a MAS model using
CORMAS (Bousquet et al.,1998) platform. This information allowed us
to generate a discussion with actors, researchers and technicians from
local institutions, and to improve the understanding of land cover
evolution in the study zone.
3.2. The Role-Playing Game (RPG)
RPG is an educative tool as well as a device that facilitates
negotiation processes in ecosystem management and it contributes
5
Satellite images utilized in cover analysis: Path Row Date Sensor Id Uni
Observations d/m/a 7 58 11-01-88 TM 032-334 Landsat 7 58 1994 TM 075-994
Landsat 7 58 03-03-01 ETMþ 042-736 Landsat.
1324 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333

Fig. 1. Location of the study zone, source, 2005. In the upper map, blue lines represent the Colombian Amazonian Region, and white lines the rest of the country’s regions. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

to the understanding of complex systems. The complexity of
natural resources management is in part built by the process of
social construction of the reality, which entails the interaction
among individual and collective mental representations about
environment, say mental models, and the context or frame that
embeds mental models (Pahl-Wostl, 2007). The discovering of
different points of view among stakeholders is not evident for
themselves in problem definition arenas for natural resources
management, and one of the tools used to understand mental
models heterogeneity are RPGs. Shared perceptions can be found
and used to facilitate coordination and negotiation processes
(Souchère et al., 2010).
 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
Fig. 2. Methodological framework.
Utilized with multi-agent systems (MAS), it allows an under-
standing and a modeling of the decision-making processes of
social actors and natural resource management (Bousquet et al.,
2001). RPGs have been used as tools for obtaining information,
as well as a means of overcoming the difficulties that imply
interviews and surveys which in many cases create barriers that
undermine the trust between the researcher and local people
(Castella et al., 2005). Therefore, RPGs have become a tool: 1) to
represent a reality about a problem; 2) to observe players
behavior; and 3) to obtain information about decisions as well as
offering a space for a discussion of the results (Castella et al.,
2005). This tool provides us with the opportunity of having
a field laboratory which permits the collective construction of
a problem representation through a participative activity, where
knowledge exchange among actors (decision-makers and
researchers) allows these to identify a solution for a given situa-
tion (Castella et al., 2005).
The decision-making rules of social actors are incorporated
into the model through the observation of the player’s behavior
(Bousquet et al., 2002; Castella et al., 2005). The game also
allows us to verify the conceptualization of the system that was
constructed from secondary information. In addition, the game
allows an understanding of the interactions among actors,
institutions and the economic aspects of the problem. Players
were selected based on the game objectives and the current
relationships between the supporting institution (SINCHI) and
the colonos population. The RPG design required us to: 1)
identify the necessary inputs for the establishment of
1325
grasslands, 2) identify land uses in order to design the game
cards that represent the soil use for each player inside their
property, 3) build a physical space to represent the territory, 4)
build game cards and pieces such as fake money, cattle pieces,
agricultural inputs cards, and 5) design recording information
protocols.
The game was designed with two types of players: active and
passive. Active players were the peasants-colonos who made deci-
sions on their property, and passive players who had no property.
Criteria for the selection of active players were: 1) belonging to the
colonos producers association, 2) living in the study zone, and 3)
living in different regions within the study zone. Six active players
participated in the RPG. Criteria utilized to choose passive players
were: 1) having experience in the assigned role, 2) being interested
in the project, and 3) being able to relate with the other active
players. Passive players were; 1) the sellers of agricultural inputs, 2)
representatives of the bank which gives agricultural credits, 3)
representative of the environmental agency that issues environ-
mental licenses, 4) technical assistants and 5) two merchants. The
institutions were chosen according to their importance to the cattle
raising activity and their influence on extraction activities on
colonos’ lands.
3.3. Interviews based on flow diagrams
The purpose of the semi-structured interview was to under-
stand colonos’ decision-making in the management of their lands.
This information became an input for the UML formalization as
1326 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
Fig. 3. Multi-temporal analysis results. Evolution of Forest and pastures and persistence of pastures between 1988 and 2001.
activity diagrams. In order to reach this objective, an interview
guide was designed based on flow diagrams for different activities
of respondents regarding the establishment of grasslands. In an
open dialog the colono was asked to describe the procedure used to
transform the land cover in his property and what the economic,
social, political and environmental factors that could drive these
decisions.
3.4. Formalization of the MAS model
In order to construct a MAS model, UML was utilized and coded
in the software CORMAS. We used Protocol Analysis methodology
to formalize the decision rules of agents, which consisted of orga-
nizing the information given by actors to be modeled (Becu et al.,
2003). To reach this objective we used the system of (Milton
1999, in Becu et al., 2003), who proposes a classification associ-
ated to semantic expressions of concepts, instances, processes,
attributes and values, and relationships. The information derived
from direct observation, interviews and role game was used to
make a list of words and expressions meaningful for the decision-
making processes of the actors. The system conceptualization
consisted in developing a class diagram and several activities
diagrams where the rules and their actions were represented.
Programming of the actions that modify the objects or elements of
the model was implemented through condition-action rules (If..
And..Then.Else.).
4. Results
4.1. Land cover change in the study zone between 1988 and 2001
Through the analysis of satellite images for the years 1988,
1994, and 2001 five land cover types were identified in the study
zone, classified as forest, stubble6, pasture, water bodies, burn
areas and no information areas for the zones covered by clouds. In
1988 the forest occupied 31.721 ha, decreasing in 1994e24.876 ha.
For the year 2001, the forest type remained the most extensive but
it covered less than half of the area of the territory. Over the 13
years, the forest cover decreased at a rate of 0.936 ha per year.
During this period (1988e2001) 12.162 ha of forest were lost. In
contrast, the pasture area increased from 7.939 ha in
1988e14.790 ha in 2001, at an annual rate of 0.527 ha per year.
During the period 1988e1994, approximately 73% of forest
remained unchanged, while 20.1% changed to stubble and 6.3% to
pasture. During the period 1994e2001, 8.3% of forest changed to
pasture and 14.7% became stubble. Finally, during the period
1988e2001, 14.2% of forest became pasture and 19.4% changed to
stubble. The annual change rates of the three main land covers
(forest, stubble and pastures) showed that the forest had the
highest change rate, becoming pasture and stubble in both
periods. Pasture increased during the whole period of the study,
and by 2001 it covered almost twice the area it occupied in 1988.
Stubble cover remained relatively stable during the studied
period.
4.2. Role-Playing Game (RPG)
RPG of land use gave relevant information about the individual
management of the properties. The game also provided
6
The Spanish term is “rastrojo” and it refers to “the lower short stiff part of the
stems of crops such as wheat, etc. that are left in the ground after the top part has
been cut and collected” Oxford Dictionary. In this study stubble is used to designate
the stubble of pasture and crops, in order to understand its spatial dynamics as
a land cover.
 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
information about the decision rules that colonos apply when they
transform land cover, as well as information on the social relations
among actors. Initial conditions of the game were set according to
the actual inventory that players made of their own property in
terms of agricultural inputs, number of cows and general equip-
ment. Players received an amount of money (fake) that they said
they actually had. The players were asked to represent their
property in the board game using game cards of their assigned land
uses. These cards had a colored border indicating each player’s
property.
The initial spatial global configuration of the properties did not
have any relation with the actual landscape because each player
came from a different region of the study zone. However, land cover
inside each one of the six properties was represented as the real
land cover that they had, so that they played from a real situation
instead of from an idealized one. At the beginning of each round,
players made decisions assuming that the season was summer
(December), and were required to change the land use cards
following real life procedures. When they needed inputs, they had
to buy them, and if they wanted to sell products they had to go to
the market. Fig. 4 portrays the board game with each player’s
representation of their property at the end of each round. Players
changed the initial land cover depending on their individual deci-
sion rules related to their monetary capital, annual activities, cattle
quantity and other factors.
Throughout the RPG, individual and collective decisions were
identified, as were the factors that drive the land cover trans-
formation dynamics. By analyzing the property dynamics, the 6
players made common decisions for the management of their
properties, and these are reflected in land cover changes and
persistence (see Table 1):
Fig. 4. Role Game results for the 4 rounds played. Due to light conditions during photography shooting, the first line of photos shows an apparent difference between red and pink
patches, but the real color is red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
1327
i) The establishment of pastures: all the players made the
decision to establish new pasture areas, because they had
the economic capacity to do this. The criteria for locating the
pastures was the potential neighborhood area of former
pastures. In many cases the farmers increased their pastured
areas in order to raise cattle at some future time.
ii) Decision to transform land cover to pasture: all the players
cultivated pastures of Brachiaria or Hyparrenia genus replac-
ing forest or stubble between December and March. Players
who had stubble prioritized slash and burn practices for
between 2 and 5 years. All the players increased their grass-
lands during the game rounds because they always had the
economic capacity due to their selling of cattle and other legal
and illegal activities.
iii) Maintaining pastures: persistence of pastures depends on the
maintenance practices. All the players carried out these
practices from 1 to 3 times per year.
iv) Establishment of other crops: all the players maintained crops
for their own and livestock consumption during all rounds of
the game. Plot size depended on household composition and
the desired level of income from corn.
v) Conserving forest plots: all players maintained forest patches
in order to supply timber for the farm, especially firewood,
fruit and construction materials. The permanence of these
patches is explained because farmers see them as a land
reserve for future areas of pasture when the older ones
become degraded.
Some players made particular decisions not common to the
group, and these individual decisions influenced the land cover
transformation:
1328 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333

Table 1
Decisions of peasant-colonos in land cover transformation towards pastures.

Decision Depends on Processes

Pastures implementation Money for land closure To verify potential area for transforming
To verify monetary resources
To change land cover and land use To have land with stubble or forest cover To verify area for cover transformation
Proper year season To verify familiar labor force or monetary resources
Monetary resources or familiar labor force
Cleaning of existing grass To have grass with weeds To verify monetary resources
To verify familiar labor force
To contract labor force Availability of familiar labor force To verify monetary resources
To buy heads of cattle Available parcels To verify available heads of cattle
To sell heads of cattle Old cattle To verify cattle condition
Monetary urgency
Young bulls between 450 and 200 kg
Number of heads of cattle is higher than available
pastures parcels shortage of pastures
To sell labor force Available time To verify monetary resources
Small size of the property
To ask for credit Need of resources for cleaning grasslands, cut forest, Double check if he has property and cattle
activities out of the property and cattle buy
To conserve the forest Need of materials from forest To verify the amount of available forest
Degraded condition of the parcels Localization inside the farm
Presence of creeks

1. Player 1 decided to enrich the stubble with woody and pioneer
species in order to support the vegetal succession dynamics
and to replace degraded soils by pasture. This was evident by
the change from stubble to forest in 1 or 2 years.
2. Management of pastures and stubble through crop rotation
schemes (player 5).
3. Forest conservation decisions were made in order to maintain
creeks and water sources in the property (players 2 and 5).
4. Player 3 did not permit stubble development because he didn’t
know how to manage stubble, therefore his decision was to
increase pastures in relation to forest.
cropping and the undertaking of seasonal work when a labor
force is needed.
7. Incentives for cattle raising. Farmers receive economic support
when they want to establish pastures, because this activity is
a guarantee for lender institutions. The establishment of
pastures becomes a requisite for credit allocation.
8. Legal character of cattle raising. As a legal productive system in
the zone, cattle raising is the most profitable activity.
9. Non-compliance of formal environmental rules. Local actors do
not follow formal environmental regulations, and environ-
mental institutions are not present in the zones.

Regarding aggregated level, when the amount of forest began

decreasing, players transformed the stubble into pasture in order to
avoid the complete loss of the forest.
During the debriefing, players agreed on the reasons for increasing
the amount of pasture lands, and identified factors such as:
1. The desire to be a stable cattle raising farmer. Players always
made decisions seeking to increase pastures, even if they had
the opportunity of increasing their income through different
legal or illegal activities or from off-farm activities.
2. Land possession and its value increase. The players preferred to
deforest rather than rent pastures on someone else’s property.
This investment in labor and the establishment of pastures
implies that the territory is not considered as vacant land.7
3. Membership of a cattle raising producers association. Players
belong to an association that gives incentives for milk and its
derivative products and commercialization at local and
national levels.
4. Existence of a market for cattle producers. There are no markets
for agricultural products.
5. Labor force availability. Unlike with agriculture, the establish-
ment and maintenance of pastures do not require high
investments in a labor force.
6. Income from other activities. Cash flow can be achieved from
activities such as trade, the provision of transport services, coca
7
The term used in Spanish is “baldio”, which means a lack of private property
rights, but also that has not been transformed. The “baldios” are territories under
the Nation domain and sensitive to private appropriation.
4.3. The model
The formalized UML system was partially coded on the COR-
MAS platform. The computer model was built with very simple
rules, where peasant-colonos only needed to change land use. The
objective of the computer model was to explore its scope and
analytical power, not for the generation of discussion spaces or
scenario building. The main objective of the construction of a MAS
model was to represent the current situation of deforestation in
the study zone as a consequence of social actors’ decisions leading
to the establishment of pastures for cattle raising in an Amazo-
nian colonization zone context. As a first step we identified the
most important entities, their relationships and processes. A
number of assumptions were made: 1) Environmental conditions,
such as precipitation and solar radiation, were not considered in
the first version because they present monthly variations whereas
decisions regarding pasture implementation are yearly and the
time frame of the model was one year. 2) Costs of the establish-
ment of each type of land cover includes: the average cost of
contracting labor according to the number of hectares the player
wants to transform, the cost of transforming land cover, and the
average costs of the required inputs for establishing perennial or
annual crops and pastures. 3) All agents (peasants-colonos) have
a market to sell their products. 4) Production and maintenance
costs are the same for all agents. 5) The closeness to the road
factor was not included in the model. In the zone, this is a factor
that directly affects the production costs. 6) The price of the cattle
is randomly generated in each time step. 7) Each agent has only
one property.
 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
Model components are classified as spatial, localized and social
entities (Table 2). The minimum spatial unit in the model is 1 ha.
Localized entities are the cattle raising farm and the types of land
cover on the land parcels. Social entities are the latifundistas, which
have one unique strategy that has to do with their social position as
large estate owners, pioneer-colono and peasant-colono. The agents
have the capacity to change the strategy according to the possi-
bilities they have had for land and capital accumulation. The actions
carried out by each type of producer are shown in Table 3. The
income generated by illegal crops (coca) was included as one of the
land covers in the process of establishing pastures. Importantly, this
type of land cover was not registered in the multi-temporal analysis
due to the scale of the satellite images. As a consequence, in the
model there are new types of land covers, which in reality are land
use units such as corn and coca. The importance of coca crops is
that they are a source of high income, and corn crops are important
as a subsistence product during the initial phases of the coloniza-
tion process.
Several activity diagrams were developed for each strategy such
as the main activities in land cover transformation. Fig. 5 illustrates
the activity diagram for the peasant-colono. This actor, if he is
the owner of a property, verifies the number of pasture hectares
he needs, based on the number of his cattle. Then he verifies the
number of hectares that he can transform. If he has sufficient
monetary capital and labor, he changes the land cover. If he does
not have enough capital, he sells heads of cattle. At the end of each
time step he sells coca if he has part of his land dedicated to that
type of crop. He pays the total costs of implementation and main-
tenance of pastures and finishes the activity. We developed the
formal basis for coding the entire UML model, which will be an
important tool for discussion among local and regional actors and
scenario building exploration.
5. Discussion and concluding remarks
The objective of the research was to understand the contextual
factors that influence the decision-making of the peasant-colonos
regarding land use; the direct decisions in land cover trans-
formation at the individual property level; and how the peasant-
colonos generate a landscape dynamics in the zone. The following
section discusses the relationships between landscape dynamics
and the findings regarding the decision-making processes, and the
implications of using tools such as RPG and MAS models, as well as
their usefulness for land use planning.
5.1. Landscape dynamics, and colono’s strategies
5.1.1. Landscape dynamics
The integration of several tools allowed us: 1) to understand and
quantify the dynamics of landscape transformation in the study
zone at the local and regional scale, as well as the mechanisms of
agricultural frontier expansion, and 2) to confirm the existence of
the three types of intervention proposed by SINCHI (1998); the
high, moderate and low human intervention zones. In the first
Table 2
Objects and elements of the model.
Spatial Social Located
Study zone Latifundista Cattle raising farm
Parcel Pioneer-colono Pasture
Farm (Individual property) Peasant-colono Corn
High intervention zone Illegal crop
Moderate intervention zone Forest
Low intervention zone Stubble
1329
Table 3
Actions carried out by the different types of colonos.
Latifundista Peasant-colono Pioneer-colono
To contract labor force in To sow illegal crops To sow illegal crops
order to maintain de farm, To sow corn To sow corn
to buy small properties To contract labor force To sell corn
with pastures To sell land To sell labor force
To buy land To sell land
To buy illegal products To by cattle
To contract labor force To sell illicit products
if he does not have To colonize
familiar labor force
period between 1988 and 1994 forest decline was greater (21.6%)
than during the second period (1994e2001) (16.8%). Similarly,
pastured land increased more during the first period (45%) than
during the second (41.4%). This behavior is related to the economic
crisis generated by the illicit crops eradication policies that
occurred during the second period. The main effect of this policy
has been that some producers ceased investing in cattle due to
a lack of cash.
Usually, major attention is put on the trends of declining forest
cover, but stubble cover also plays a key role in the establishment of
pastures, and in general in the cover change process. Its importance
lies in the fact that the producers clear the stubble first, maintaining
the forest as a source of goods and services in the process of
establishing pastures. The areas covered with stubble are used for
corn, sugar cane, yucca, plantain, rice and pastures crops. Eventu-
ally, when the forest reserve begins to decrease, the stubble is the
space for new land uses. When the stubble disappears, the forest is
cut again.
The trend of land cover transformation in the board game, in
which four rounds representing one year each were carried out,
represents the final phase of the dynamics analyzed by Etter (1992)
and Karremans (1990) for the farms in the Guaviare colonization
zone. According to Etter (1992) the change pattern of an average
farm presents a trend of forest decline until they disappear over
a 20-year period. Crops tend to increase during the first 3 years of
the colono’s arrival, after which they decrease and stabilize. The
stubble area increases during the first 10 years of the colono’s
arrival and then decreases until it reaches a very small size or
disappears altogether, while pastures increase. According to
Karremans (1990), in the 100 ha farms, the forest has disappeared
after 30 years. Pastures increase linearly from the second year of
occupation until they occupy all of the farms after 40 years. Crops
occupy almost the same area and the areas of stubble tend to
increase at first, but after 30 years it quickly decreases until it finally
disappears.
5.1.2. Peasant-colonos’ decisions
Table 4 shows the driving factors that have brought the system
to its current state. These factors, identified through the historical
analysis, interviews and role game, contextualize and guide the
individual decisions of peasant-colonos in the management of their
farms. Traditionally, scholars and institutions have claimed that the
colonos, in the first phases of settlement, develop non-sustainable
practices in the management of natural resources and land use
(Bakker, 1993). But as time passes, they improve their practices and
adapt to environmental conditions and thus obtain better results.
Decisions change over time, and even if the area under pasture
declines, they introduce conservation strategies to the remaining
forest patches and water sources. The context of the illegal
economy in which the region lives can be seen in: 1) the natural
resources extractive economy, 2) the colonization surges caused by
violence in other parts of the country, 3) certain governmental
1330 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
Fig. 5. Activity diagram of the type of actor peasant-colono.
policies, 4) the concentration of land ownership, 5) high production
costs, and 6) environmental constraints. The illegal economy has
been used to satisfy the producers’ economic needs, and they have
thus become a part of the cattle raising model.
Table 4
Relevant driving factors for pastures implementation.
Driving factors Type
Socio cultural factors Agrarian tradition
Cattle raising system experience
Social and productive model: latifundista
Land property
Land possession
Land property concentration
Social organization
Economic factors Market for cattle raising activity products
Lack of market for agricultural products
Increase of heads of cattle
Rents due to cattle raising
Familiar labor force availability
Revaluation of the land after deforestation
Cattle raising cycles
Income from other sources
Institutional factors Incentives for cattle raising activities
Big distances between environmental law
and social actors
Land property title issuing
Technological factors Training for cattle raising activities
Access to cattle raising technology
Access to technology for slash and burn
activities
Factors related with Environmental territorial offer
environmental territorial offer, Distance to main roads
accessibility and time Distance to human settlements
Time needed for establishment in the zone
The topic of illegal crops appeared only in the RPG. Players gave
concrete information about the fact that illicit crops allowed them
to generate higher revenue than other legal products. In addition to
that, cattle raising was found to be a safe income source over time.
Legal agricultural products produce low revenues due to infra-
structure deficiency, large distances from human settlements and
lack of markets, therefore the main legal productive activity is
raising cattle. The analysis of illicit crops dynamics is a very sensi-
tive topic. Among the findings of the study is the fact that income
generated by other sources (illicit crops, services and trade) are
essential for the establishment of pastures. Production system
diversification (growing crops for personal consumption and cattle
raising) works as a buffer for the crisis periods of coca price, which
is the product that yields the highest economic benefit.
We were able to observe the relationships between colonos and
the institutions present in the zone. An aversion to availing of bank
credit was evident due to the high interest rates and the bureau-
cratic procedures. However, a flow of capital occurred among active
players (colonos) outside the formal channels, such as the financial
institutions, because of the revenues from illegal crops, which flows
informally through the system. Colonos did not ask for environ-
mental licenses, a fact that was mentioned in the debriefing where
the discussion was about the problems of deforestation and the
difficulties of compliance to the formal rules. Players were allowed
to make agreements amongst themselves, and with institutions, as
well as working out of their properties and carrying out any other
type of activity. This fact provided evidence of the off-farm activi-
ties and its importance to the household economy which is not
exclusively based on land work. Representatives of institutions as
passive actors represented their actual role. It was possible to
observe the distance between the small and middle land owners
 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
with banks, as well as with environmental and technical assistance
institutions. Decisions about land use planning at the individual
farm scale followed the same general strategy, which was imple-
mented by all active players independently. This behavior
confirmed the information gained from personal interviews.
As one of the important lessons from the study, we can say that
the generalized idea that the peasants who produce coca are
delinquents is an extremely simplistic interpretation. This activity
is carried out by people pushed by difficult structural conditions, in
many cases generated by the state, towards the peripheral zones.
Agrarian policies and governmental low capacity for rural conflict
management have triggered several effects: high land property
concentration, reduction of productive areas, and employment loss
in the agricultural sector among others. Under this scenario, the
unique alternative left for peasants has been migration towards
the cities or frontier marginal zones (Fajardo, 2002). In these zones
the complex context driven by the precarious presence of the state,
and a strong domination of armed groups, the coca production is
the only means available to them to improve their livelihoods.
While in terms of land use importance and environmental negative
effects, coca cropping plays a minor role, economically and socially
it is an important activity.
5.2. Future perspectives
The approach taken in this study unveils the logics of illegal
agricultural practices at the micro scale. It contrasts the dominant
vision about colonos as delinquents against a vision that attempts to
understand the dynamics and logic of land use management in
these contexts. An important finding is that the illegal coca crop-
ping is a fundamental component of the livelihood system, and to
change this will require innovative approaches. The micro economy
of coca cropping and the land use management logic in Colombian
colonization zones have not been studied in depth. Torres (2000),
Polícia Nacional de Colombia-Dirección Antinarcoticos (2004) and
Caicedo (2006) report partially on the micro economic dynamics of
illegal coca cropping, but still there is a lack of a detailed and
integral understanding of the micro dynamics of the phenomenon.
We claim that the use of this type of methodological approach, in
contexts such as Guaviare colonization zone, is useful in guiding
programs and testing scenarios including government policies in
areas such as illegal crop eradication, substitution of these products
and incentives for forest conservation. The approach could inte-
grate not only ecological and economic dimensions, but the social
and human dimension from the point of view of colonos in scenario
analysis type of projects.
5.3. Methodological tools and polysemic objects: talking tools
The research confirmed the effectiveness of the methodological
approach for understanding territorial logics of peasant-colonos.
Traditional ethnographic methods can be used to carry out this type
of study, but they imply several problems: 1) in depth interviews
and direct observation implies long time periods in the field, and
a long process of building trust. Usually people will not freely
discuss topics such as their illegal activities or forest conservation,
which, in this study, are fundamental to understanding micro
decisions of land management and landscape dynamics. 2) feed-
back to local communities from the latest research is not efficient.
Outputs are not easily readable and understandable by local actors.
The consequence is that people do not engage in a self-reflection,
and they usually view researchers as people who gather informa-
tion. And 3) there is a sort of tiredness in local communities
regarding traditional participatory and ethnographic methods,
where people systematically reject participating in projects. These
1331
circumstances make the approach, which we carried out, mean-
ingful to both locals and researchers.
From our perspective, to perform research in regions such as the
colonization zone of Guaviare needs innovative approaches for the
following reasons: 1) The area presents historical natural resource
exploitation dynamics, which have increased the ecological vulner-
ability of the Amazonian ecosystems. 2) They provide the possibility
of establishing direct interaction with local actors, without talking
about “hot” topics, such as coca cropping and forest conservation. 3)
The need for representing spatially the agricultural expansion allows
the observation of space occupation dynamics and natural resource
use within the farming community.4) They provide the possibility of
understanding and finding the key components, structure and
relations of a globally complex phenomenon. 5) Innovative
approaches allow the opportunity to trigger a local process of self-
reflection regarding territorial dynamics at the individual level. This
capability of a community is fundamental for territorial planning and
natural resource management, concerning for example, the negoti-
ation of forest areas requiring conservation, the possibilities of
switching from illegal crops to alternatives for capitalization. And, 6)
they provide the potential to simplify complexities, to gather an
important amount of detailed information at the same time as giving
feedback to participants.
The interviews, based on activity diagrams, were fundamental
to understanding actors’ decisions in vegetal cover transformation
at the individual scale. These diagrams facilitated the construction
of UML dynamic diagrams, integrating the points of view of
peasant-colonos and researchers. The construction of activity
diagrams with colonos constituted a solid verification source for the
final UML formalization. The Role-Playing Game generated
a participative discussion in a conflict region, which was used to
obtain information without visiting the different sub regions of the
zone, to verify social, economic and environmental dynamics and to
identify interactions among players. During the game, limitations
and potentialities were identified. The principal difficulty during
the game was recording all of the information because of the
synchronicity of events and the high amount of information
generated. Although players were accompanied by a facilitator who
helped them record the information in the correct format, it
remained difficult to record all the information such as the
communications among players. Barreteau et al. (2001) recom-
mended that this type of game adjust to what is intended to
represent of the real system, and whether the spaces can be
separated or not, materials must facilitate players to quickly enter
into the dynamics, and finally the duration of the process, if it is
complex, compels to separate the learning sessions of those related
with the game. The role-playing game of this study was developed
in one space, materials let participants respond quickly to the
dynamics. However, the time available was not sufficient to play the
desired number of rounds. It is recommended to carry out another
game session with pioneer-colonos and latifundistas in order to
understand the interactions among all the actors. This would be
relevant in the territorial planning of the region. In this case, the
final discussion would give useful insights, having taken into
account natural resource conflicts among the different types of
actors present in the region.
MAS modeling was used as a device that permits the researcher
to explore the limits of simplification of the systems. As we already
mentioned, the model used was a very simple prototype that could
reproduce the general trends of landscape transformations. A
gradual increase in model complexity will give insights into the
essential mechanisms responsible for the studied behavior. When
MAS modeling is accompanied with RPG, it becomes a tool for
understanding and monitoring negotiation processes and the study
of complex systems (Barreteau et al., 2001). In this study, these
1332 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
tools were used for the second purpose: understanding the system.
Negotiation and monitoring aspects were mentioned in the
debriefing by representatives of environmental institutions as
a future potential use. The construction of the conceptual model in
UML was carried out in a permanent feedback process which, after
an initial version based on secondary information, was com-
plemented and simplified with interviews, field observation and
RPG information.
The implementation of the RPG left us with important lessons
regarding the way the tool worked, and the multiple meanings for
the participants, including researchers. The RPG constitutes
a device that could “talk” by itself, in the name of local actors.
Facts that hardly could be communicated in an interview emerge
implicitly and explicitly through the exercise. As mentioned above,
issues such as the role of coca cropping and forest clearing and
conservation in land management, and the logics of economics are
not easily understood with other methods. Therefore, we claim
that the RPG works as an artefact in the sense of Suchman et al.
(2002). Insights from RPG implementation are generated by
several points of view and agents’ interests. RPG offers a clear
“case of a performative artefact that works to align multiple,
discontinuous social worlds. Like any technology, the prototype does
not work on its own, but as part of a dynamic assemblage of inter-
ests, fantasies and practical actions, out of which new socio-material
arrangements arise.” (Suchman et al., 2002, 175). When the RPG is
implemented it becomes an object that is different and brings
new meanings for each actor “objects and actions are realized
within culturally and historically reiterated fields of possibility, which
afford the familiar ground of everyday experience.” (Suchman et al.,
2002, 175).
In a further phase of the project, consisting in MAS model
presentation to actors and participative simulations, both RPG and
the MAS model could work as polysemic objects that “speak in
different voices for different audiences” (Suchman et al., 2002, 174).
They become talking objects that address local actors and
researchers.
The RPG is a device that we call a “dense methodological tool”,
in the sense that it is a designed object that synthesizes a complex
system. It is dense because there are a number of variables, rela-
tions, conflicts, roles, rules, time and space compressed in one
comprehensive representational object. In this logic, we can think
of the MAS model as a meta-object where time and space
condensation is higher. The MAS model is a simplification of the
complex system, reality and RPG, but this simplification implies
a meaningful abstraction of a higher degree than the RPG. From this
perspective, RPG and MAS models are relevant because they help
local actors observe their “space of experience” and their “horizon
of expectations” (Kosellek, 1993) in a different way. According to
Kosellek, the “space of experience” refers to the individual and
collective history in which multiple time strata are compressed
without sharp differences of their before and after. This is not an
experience measurable chronologically. The “horizon of expecta-
tions” is constituted by the line that marks the future and the
beginning of a new space of experience. These concepts are central
to territorial planning. RPG and MAS models talk to actors and
researchers in the same language that human memory and
projection mental capabilities function. These objects condense
time and space and allow the making of problems explicit, finding
solutions and exploring possible scenarios.
Acknowledgements
We want to express our gratitude to all persons who partici-
pated and made possible this project in Bogotá, San José del
Guaviare, Brasilia and Montpellier. Special thanks to all the colonos
that participated in this work, to people from the SINCHI institute in
Guaviare and Bogotá, to Guillermo Vargas, Mauricio Zubieta,
Bernardo Lucena and Gladis Gavilán, Uriel Murcia, Carlos Ariel
Salazar and the people from institutions in San José CDA, ICA and
CINDAP. To Diana Maya, from Javeriana University in Bogotá, for her
interest and invaluable help during field work, to Cristophe Le Page,
Pierre Bommel and François Bousquet from CIRAD for their help
and the openness in sharing their modeling knowledge and skills.
Thanks to Jean François Tourrand from CIRAD for his constant
interest and for providing us the resources for traveling to Brasilia
and Montpellier. Special thanks to Gloria Inés Restrepo who gave us
key comments and suggestions during the writing of the paper.
Finally, we acknowledge the work of three anonymous reviewers
who read the paper and gave relevant suggestions for the writing of
this article.
References
Bakker, L.M., 1993. Colonization and Land Use in the Humid Tropics of Latin
America. Werdocument IKC-NBLF NR. 34. Foundation for Netherlands Forestry
Development Cooperation, Wageningen, Netherlands.
Barreteau, O., Bousquet, F., Attonaty, J.M., 2001. Role-playing games for opening the
black box of multi-agent systems: method and lessons of its application to
Senegal River Valley irrigated systems. Journal of Artificial Societies and Social
Simulation 4 (2). http://www.soc.surrey.ac.uk/JASSS/4/2/5.html.
Becu, N., Bousquet, F., Barreteau, O., Perez, P., Walker, y.A., 2003. A methodology for
eliciting and modeling Stakeholders representations with agent based
modeling. In: Hales, D., Edmonds, B., Norling, E., Rouchier, J. (Eds.), Multi-agent-
based Simulation iii. 4th International Workshop, MABS 2003 Melbourne,
Australia, Revised Papers. Lecture Notes in Artificial Intelligence 2927. Springer,
pp. 131e148.
Bousquet, F., Barreteau, O., d’Aquino, P., Etienne, M., Boissau, S., Aubert, S., Le
Page, C., Babin, D., Castella, J.C., 2002. Multi-agent systems and role games: an
approach for ecosystem co-management. In: Janssen, M.A. (Ed.), Complexity
and Ecosystem Management: the Theory and Practice of Multi-agent Systems.
Edward Elgar, Cheltenham, UK, pp. 248e285.
Bousquet, F., Trébuil, G., Boissau, S., Baron, C., d’Aquino, P., Castella, J.C., 2001.
Knowledge integration for participatory land management: the use of multi-
agent simulations and a companionable modelling approach. Chiang Mai,
Thailand. In: Participatory Technology Development and Local Knowledge for
Sustainable Land Use in Southeast Asia. http://cormas.cirad.fr/en/bibliog/
communi.htm.
Bousquet, F., Bakam, I., Proton, H., Le Page, C., 1998. Cormas: common-Pool
Resources and Multi-Agent Systems. Lecture Notes in Artificial Intelligence
1416, 826e837.
Caicedo, C., 2006. Expectativas racionales y persistencia de cultivos ilícitos de coca
em Colombia. 1986e2003. Bogotá. Trabajo de grado publicado para optar al
título de Magister en Planificación y Administración del Desarrollo Regional.
CIDER. Universidad de los Andes.
Castella, J.C., Trung, T.N., Boissau, S., 2005. Participatory simulation of land use
changes in the Northern Mountains of Vietnam: the Combined use of an agent-
based model, a role-playing game, and a Geographic information system. Ecology
and Society 10 (1), 27. http://www.wcologyandsociety.org/vol10/iss1/art27/.
Etter, A., 1992. Mapa ecológico General de Amazonía Colombiana (escala
1:1’500.000). In: Andrade, G., Hurtado, A., Torres, R. (Eds.), Amazonía
colombiana: diversidad y conflicto. CEGA-COLCIENCIAS, Bogotá.
Fajardo, D., 2002. Para sembrar la paz hay que aflojar la tierra. Comunidades, tierras
y territorios en la construcción de un país. Instituto de Estudios Ambientales
IDEA. Universidad Nacional de Colombia. UNIBIBLOS, Colombia.
Forero, J., Guevara, L.E., Ortiz, P., Durana, C., Galarza, J.A., Corrales, E., Rudas, G., 2002.
Sistemas de producción rurales en la región andina colombiana: análisis de su
viabilidad económica, ambiental y cultural. Facultad de Estudios Ambientales y
Rurales, Pontifica Universidad Javeriana, Conciencias, Bogotá.
Karremans, J.A.J., 1990. El proceso de colonización en el Guaviare, un análisis
cuantitativo. Instituto de Antropología, Universidad de Utrecht, Holanda, Cor-
poración Araracuara, Colombia.
Kosellek, R., 1993. Futuro pasado. Para una semántica de los tiempos históricos
Barcelona, Paidos.
Milton, N., Shadbolt, N.R., Cottam, H., Hammersley, M., 1999. Towards a Knowledge
Technology for Knowledge Management. International Journal of Human-
Computer Studies 53 (3), 615e664.
Murcia, U.G., 2003. Análisis de los procesos de deforestación y praderización en las
zonas de colonización de la Amazonia colombiana. Estudio de caso Departa-
mento del Guaviare. Periodo 1987e2001. Tesis para optar por el título de
Magíster en gestión Ambiental para el Desarrollo Sostenible. Pontificia Uni-
versidad Javeriana, Facultad de Estudios Ambientales y Rurales, Bogotá,
Colombia.
Ordónez, N., López, E., 1989. Impacto de la colonización en la degradación de los
suelos., torcha oriental, zona de colonización del Guaviare. COA, Programa
 M. Vieira Pak, D. Castillo Brieva / Environmental Modelling & Software 25 (2010) 1322e1333
Guaviare, Unidades de suelo, Proyecto Colonización. Universidad Utrecht,
Colombia.
Palacios, G., 2001. En búsqueda de conceptos para una historiografía ambiental.
Página 39 en Naturaleza en Disputa. In: Ensayos de la Historia Ambiental de
Colombia 1850e1995, Primera Edición. Unibiblos, Bogotá, Colombia.
Pahl-Wostl, C., 2007. The implications of complexity for integrated resources
management. Environmental Modelling & Software 22, 561e569.
Polícia Nacional de Colombia -Dirección Antinarcoticos, 2004. Acercamiento a la
actividad agronómica y la problemática ambiental de los cultivos de coca.
Bogotá. D.C. Septiembre.
Salazar, C., 2002. Los asentamientos rurales en zonas de colonización: el caso de la
Amazonía occidental. Tesis de Maestría en desarrollo rural. Pontificia Uni-
versidad Javeriana, Bogotá.
1333
SINCHI, 1998. Diagnóstico de los actuales sistemas de producción y su impacto
ambiental en la zona de colonización del Guaviare. Instituto Amazónico de
Investigaciones científicas.
SINCHI, 1999. Guaviare: Población y Territorio. Tercer Mundo Editores, Bogotá,
Colombia.
Souchère, V., Millair, L., Echeverria, J., Bousquet, F., Le Page, C., Etienne, M., 2010.
Co-constructing with stakeholders a role-playing game to initiate collective
management of erosive runoff risks at the watershed scale. Environmental
Modelling & Software 25 (11), 1359e1370.
Suchman, L., Trigg, R., Blomberg, J., 2002. Working artefacts: ethnomethods of the
prototype. British Journal of Sociology 53 (2), 163e179.
Torres, L.E., 2000. Costos de Producción de coca a nivel campesino en la región del
caguán, Caquetá e Colombia. Un análisis comparativo Bogotá.