You are on page 1of 13

Reg Environ Change

DOI 10.1007/s10113-013-0433-7

ORIGINAL ARTICLE

National and regional determinants of tropical deforestation


in Colombia
Dolors Armenteras • Edersson Cabrera •

Nelly Rodrı́guez • Javier Retana

Received: 29 August 2012 / Accepted: 25 February 2013


Ó Springer-Verlag Berlin Heidelberg 2013

Abstract Global tropical deforestation continues to occur primarily due to fires related to the colonization front.
at high rates despite political attention. National-level for- Policy mechanisms will have to take into account regional
est baselines are being established all over the world to patterns to successfully balance development and forest
guide the implementation of several policy mechanisms. preservation in Colombia.
However, identifying the direct and indirect drivers of
deforestation and understanding the complexity of their Keywords Tropical forest  Deforestation  Drivers 
interlinkages are often difficult. We first analyzed defores- Heterogeneity  Regional  Colombia
tation between 1990 and 2005 at the national level and
found an annual deforestation rate of 0.62 %. Next, we
performed separate analyses for four natural regions in Introduction
Colombia and found annual deforestation rates between
0.42 and 1.92 %. Using general linear models, we identified Global forest loss rates are reported at 0.6 % per year for
several direct causes and underlying factors influencing the 2000–2005 period (Hansen et al. 2010), with an esti-
deforestation at the national level: rural population density, mated area of global forest cover loss between 2000 and
cattle, protected areas, and slope. Significant differences in 2005 of 1,011,000 km2. The destruction of tropical forests
deforestation rates and causes were found across regions. In remains a critical environmental issue given the unique
the Caribbean region, drivers of loss are urban population, ecosystem services these forests provide and the high
unsatisfied basic needs, slope, and precipitation and four biodiversity they host. Of all forest types, the humid
land use variables (illicit crops, pastures, cattle, and fires). tropical forests cover the largest area (approximately
In the Orinoco region, crops are the main driver of forest 11,564,000 km2 in the year 2000), but these are being
loss, and in the Amazonian region, deforestation is destroyed at an average rate of 2.4 % of forest cover per
year (Hansen et al. 2010). Loss rates differ around the
world and rates are inconsistent throughout the tropical
D. Armenteras (&)  N. Rodrı́guez moist forest biome. Achard et al. (2002) reported annual
Laboratorio de Ecologı́a del Paisaje y Modelación de
deforestation rates between 1990 and 1997 of 0.38 % for
Ecosistemas, Departamento de Biologı́a, Facultad de Ciencias,
Universidad Nacional de Colombia, Edificio 421, Latin America, 0.43 % for Africa, and 0.91 % for South-
Bogotá D.C, Colombia east Asia as well as a global rate figure of 0.52 %.
e-mail: darmenterasp@unal.edu.co A similarly variable picture emerges when examining
the causes and main agents that drive tropical deforestation
E. Cabrera
Proyecto REDD, Instituto de Hidrologı́a, around the globe (Rudel and Roper 1997; Geist and
Meteorologı́a y Estudios Ambientales (IDEAM), Lambin 2002; Rudel 2006; Kindermann et al. 2008; Carr
Carrera 10 No. 20-30, Bogotá D.C, Colombia et al. 2009). When analyzed from a social science per-
spective, population pressures (as an underlying driving
N. Rodrı́guez  J. Retana
CREAF i Unitat d’Ecologı́a, Universitat Autònoma force or indirect driver) and shifting cultivation (as a
de Barcelona, 08193 Bellaterra, Barcelona, Spain proximate cause or direct driver) are often identified as the

123
D. Armenteras et al.

main factors behind deforestation (Geist and Lambin et al. 2007) and many of those studies that are at the
2002). In 2001, Latin American population had doubled national level (Herrador Valencia et al. 2011; Angulo 2011;
due to a decline in mortality, improved health and Huang et al. 2009; Wright and Samaniego 2008) do not
economic development (Carr et al. 2009). A cross-national incorporate differences among regions and thus do not take
study of deforestation causes (Rudel and Roper 1997) into consideration how climatic drivers vary across the
concluded that tropical deforestation was better explained country or affect deforestation more substantially in some
by the interaction of multiple factors, among which eco- areas, whereas human-related drivers predominate in oth-
nomic forces were predominant, followed by infrastructure ers. These studies also often do not look at the interlinkages
and road development. In Latin America, demographic among direct and indirect drivers, and some authors argue
factors associated with population growth with lower rural that countries remain relatively cohesive socioeconomic
fertility and high rates of both rural-to-urban migration and units (Rudel and Roper 1997). For Colombia, some
also rural-to-rural migration pushing colonization fronts regional similarities in deforestation drivers were proposed
further into the forest have shown an important influence suggesting a similar effect across regions of soil, accessi-
on deforestation (Geist and Lambin 2002; Grau and Aide bility, and forest and secondary vegetation (Etter et al.
2008; Carr et al. 2009). Most of the forest loss to agri- 2006). However, this study was based on a model of
culture in Latin America has already been done, and recent deforestation build from one single time step and left out
forest conversion mostly occurs in marginal lands (Carr several important factors such as population density. In this
et al. 2005). Recent studies have also emphasized the paper, we assess the drivers of deforestation at the national
increasing global food demand (particularly in Southeast level in Colombia, a megadiverse country with two global
Asia) (Grau and Aide 2008) and urban growth as a major biodiversity hotspots that is currently facing many devel-
driver in the twenty-first century (Defries et al. 2010). opment challenges (Jarvis et al. 2010), experimenting rapid
Environmental factors, including geology, topography, economic growth (Cotte Poveda and Pardo Martı́nez 2011)
and soil quality, and climatic factors, such as drought, are and with current processes driving deforestation that
also believed to strongly affect deforestation (Geist and include illicit cultivation (Dávalos et al. 2011). Taking
Lambin 2001; Chomitz and Thomas 2003). However, most into account Colombia’s biophysical, socioeconomic, and
deforestation studies that mention environmental factors demographic heterogeneity, we also explore the sub-
focus primarily on soil-related features (Fearnside 1984; national variation in deforestation and its causes within the
Geist and Lambin 2001). In Amazonia, the intensity of the country’s four main natural lowland regions.
dry season is one of the most important environmental
factors influencing deforestation (Laurance et al. 2001).
Similar results have been found in Bolivia, where drier Methods
forests were more prone to deforestation than wetter ones
(Steininger et al. 2001). Although it is currently recognized Study area
that multiple factors drive deforestation, these factors must
not only be identified, but their interactions must also be Colombia is located at the northwestern tip of South
analyzed to fully understand the causes of deforestation in America (centered at 12.2676 North, 4.135° South, 66.509°
a given country (Geist and Lambin 2002). More recent East, and 79.0 255° West). It is the only South American
studies (Rudel 2007; Butler and Laurance 2008) suggest country to have coasts on both the Pacific Ocean
that whereas some common patterns are detected, a mul- (1,350 km) and the Caribbean Sea (over 1,600 km)
tiple-factor approach that also considers variation over time (Fig. 1a). It is the fourth largest country (1,141,748 km2) in
is required (Rudel 2007). South America after Brazil, Argentina, and Peru. Whereas
For global studies, the country-level unit of analysis the western part of Colombia is mostly mountainous (45 %
might be adequate. However, national-level analyses of of the territory) and densely populated, a major part of the
rates and drivers of deforestation (both direct and indirect) country consists of plains lying below 500 m. Colombia’s
might oversimplify reality in many tropical countries, environment is dominated by tropical rainforest in the
leading to political responses that might be applicable Andes, Amazon, and Pacific basins (52.2 %), followed by
according to local conditions in some parts of the territory savannah in the Orinoco basin (13.9 %), cropland in the
but not in others. Many of the recent deforestation studies Andes, grassland, and shrubland in the Caribbean basin but
in Latin American countries are focused on specific regions also present in the rest of the territory (9 %) (IDEAM et al.
within these countries (Redo et al. 2009; Calvo-Alvarado 2007). The territory can be divided into six large natural
et al. 2009; Portillo-Quintero et al. 2012; Monzón-Alva- regions based on topography, climate, and soil type: the
rado et al. 2012; Stevens et al. 2011; López and Sierra Caribbean (Atlantic) littoral region, the Pacific region
2010; Zeledon and Kelly 2009; Forrest et al. 2008; Huang (including the Chocó Biogeographic rainforest), the

123
Tropical deforestation in Colombia

Orinoco region (including the plains of the Orinoco River an extremely wet area (rainfall over 10,000 mm/year) in
Basin, a large savannah lying in the basin of the Orinoco the western part of the country (Armenteras-Pascual et al.
River), the Amazonian region (including the Amazonian 2011). The Amazonian region contains the largest extent of
forests), the Andean region (not covered in this study forest in Colombia and is sparsely populated, but it has one
because there are some recent analysis of deforestation of the most active colonization fronts in the country
patterns and drivers that widely discussed it; Armenteras (Rodrı́guez et al. 2011). The Orinoco is a scarcely popu-
et al. 2011; Rodrı́guez et al. 2011), and also the Insular lated, relatively flat region that is dominated by grasslands
region neither covered in this paper. and pastures with a few areas of small-scale agriculture and
The four regions considered here, that is, the Amazonia, a considerable gallery forest associated with its extensive
the Caribbean, the Orinoco, and the Pacific exhibit differ- river network. The Caribbean region is a highly trans-
ent ecologies, flora, and fauna and have been subjected to formed region dominated by pastures with few remnants of
different uses (IDEAM et al. 2007). They differ in climate natural vegetation and only a few patches of dry forest.
as well, with the Orinoco and Caribbean regions being Finally, the Pacific region has a humid to superhumid
relatively dry (dry season rainfall \50 mm per month), the tropical climate and is considered one of the wettest
Amazonian region being wetter with an annual precipita- regions of the world with annual precipitation ranging from
tion over 2,000 mm/year and the Pacific region comprising 4,000 mm up to more than 10,000 mm in some locations

Fig. 1 a Geographical
characteristics of Colombia.
b Natural regions and land
cover of Colombia

123
D. Armenteras et al.

Fig. 1 continued

(Fig. 1b). This region is also sparsely populated and is sensing data in the form of over 240 Landsat multi-
globally recognized as one of the world’s most biologically spectral satellite images made using Thematic Mapper
and culturally diverse areas, containing indigenous and (TM, missions 4 and 5) and Enhanced Thematic Mapper
Afro-American reserves (Mittermeier et al. 1998; Myers Plus (ETM?, mission 7) from 1990 to 2005. The Landsat
et al. 2000). Colombia currently has 1,103 municipalities data were originally orthorectified using national
that have an area ranging from the 1,557 ha the smallest standards (MAGNA SIRGAS, Datum SIRGAS, Spheroid
(Sabaneta, Andes) to the largest with 6,556.000 ha (Vich- 1980, latitude of origin 4°3500 46.32150100 N, and longitude
ada, Amazonia). of origin 74°040 39.028501W). A semi-automated digital
classification protocol was carried out in which an auto-
Deforestation dataset mated tool (CLASliteÒ) was integrated with the tradi-
tional approach (supervised classification into ERDAS
National forest/non-forest data for 1990 and 2005 were Imagine (Leica Geosystems 2005) along with on-screen
obtained from the official project ‘‘Scientific and institu- visual interpretation. The images with minimum cloud
tional capacity building to support Reducing Emissions cover (less than 10 %) available were used for the con-
from Deforestation and Degradation (REDD) projects in struction of the 1990 and 2005 forest maps, which were
Colombia’’ (IDEAM 2011). This project used remote scaled to 1:100,000.

123
Tropical deforestation in Colombia

We calculated the rate of deforestation of each munic- municipalities. For fire data, we used the fire hotspot series
ipality at the national or regional level (Rm, in %) based on for Colombia from the Moderate Resolution Imaging
the forest maps as follows (Fearnside 1993): Spectroradiometer (MODIS) Collection 5.1 active fire
Rm ¼ ððA1990m  A2005m Þ=ðFA1990  tÞÞ  100 dataset (Davies et al. 2009). Spatial annual data on illicit
crops were provided from the yearly global illicit crop
where A1990m and A2005m are the total forest areas within monitoring program of the United Nations Office on Drugs
the municipality in 1990 and 2005, respectively, and t is the and Crime, represented in Colombia by the Integrated
time interval in years (in this study, t = 15). FA1990 is the System for Illicit Crops Monitoring project, or SIMCI
total forest area in 1990 either throughout Colombia or in (Sistema Integral de Monitoreo de Cultivos Ilicitos
each of the natural regions considered (obtaining different (UNODC 2009).
rates for each of them). The sum of Rm across all munic-
ipalities equals the total deforestation rate (either for the Statistical analyses
region considered or for the whole of Colombia).
General linear models (GLMs) were used to explore the
Datasets used for independent variables relationships between deforestation and the different
demographic (urban population and rural population),
Of the 1,103 municipalities of Colombia, we filtered out socioeconomic (unsatisfied basic needs), land use (crops,
those belonging to the Andes (as deforestation drivers pastures, illicit crops, mining, fires, and protected areas),
specific for the Andes have been explained and discussed and physical environmental variables (road density, slope,
in detail elsewhere; Armenteras et al. 2011) and those for water scarcity in dry years, temperature, and precipitation).
which forest cover data were unavailable for more than GLMs go a step beyond the multivariate regression model
20 % of the area, leaving 315 Colombian municipalities in by allowing for linear transformations or linear combina-
the final analysis (15 for the Pacific, 41 for Orinoquia, 204 tions of multiple dependent variables, including both
for Caribbean, and 55 for Amazonia). The main groups of quantitative and qualitative variables (Halekoh and
deforestation contained 15 variables (Table 1) selected as Højsgaard 2004). We specified four different GLM models
possible explanatory factors at the regional level repre- for deforestation: one for the total territory of Colombia,
sented demographic, socioeconomic, land use, and physical and one for each one of three natural regions within the
environmental features (both abiotic and biotic). The country: Amazonia, Orinoquia, and Caribbean. The model
variables used were integrated in a GIS database using the for the Pacific region was not computed given the low
municipality as the basic spatial analysis unit. When we number of municipalities with data available in this region
conducted the analysis at the regional level, we included all (N = 15). All parameters were estimated by maximum
municipalities with more than 50 % of their territory in the likelihood. To normalize the data, several variables (urban
region being considered. population, rural population density, crops, pastures, cattle,
The demographic and socioeconomic information we illicit crops, and protected area) were log-transformed.
compiled included data on unsatisfied basic needs and the Statistical analyses were carried out using STATISTICA
number of inhabitants (rural and urban). These data were 6.0 (Statsoft, 2001).
obtained from the 1985, 1993, and 2005 population cen-
suses (Departamento Administrativo Nacional de Estadı́s-
tica-DANE). Data on crops, cattle, pastures, and mining Results
were obtained from the Geoportal of the Agustin Codazzi
Geographic Institute (Sistema de Información Geográfica National and regional deforestation rates in Colombia
para la Planeación y el Ordenamiento Territorial Nacional
SIGOT, Instituto Geográfico Agustı́n Codazzi—IGAC According to the forest/non-forest map for the whole of
2011). Digital maps (1:500,000) of national protected Colombia, forest loss between 1990 and 2005 totaled
areas, indigenous reserves, municipalities, road networks, 5,116,071 ha (Fig. 2), which indicates a deforestation rate
and human settlements were also obtained from the Agu- of 341,071 ha/year. The national figure for this 15-year
stin Codazzi National Institute of Geography (IGAC 2005). period represents forest losses of 1,886,769 ha (3.9 %) in
IGAC also provided us with the digital elevation model Amazonia, 753,893 ha (4.99 %) in the Caribbean region,
based on the Shuttle Radar Topography Mission (SRTM, 507,337 ha (3.24 %) in the Orinoco region, and 472,863 ha
90 m resolution) from which a map of slope (in degrees) (6.32 %) in the Pacific region (Fig. 3). Colombia as a
was derived. Climatic variables such as the annual pre- whole exhibits an annual deforestation rate of 0.62 %. The
cipitation were derived from WORLDCLIM (Hijmans annual rate varies considerably among regions, with values
et al. 2005) and averaged for the area of each one of the of 0.42 % in the Pacific region, 0.49 % in Amazonia,

123
D. Armenteras et al.

0.83 % in the Orinoco region, and 1.92 % in the Caribbean


region. Differences in the annual deforestation rate per
municipality (Fig. 4) in the four natural regions were sig-
nificant (ANOVA test, P \ 0.01), with an average annual
deforestation rate per municipality of 0.0090 % (±0.0016)
in the Caribbean region, 0.0091 % (±0.00016) in
Amazonia, 0.0204 % (±0.00552) in the Orinoco region,
and 0.0285 % (±0.00914) in the Pacific region.

National and regional drivers of deforestation


in Colombia
Fig. 3 Total deforestation (in thousand ha) per region (black bars)
The best combination of explanatory variables used in this and the percentage of each region deforested (gray quadrats and
lines) between 1990 and 2005 in the four natural regions of Colombia
paper of deforestation for Colombia at the national level
(R2 = 0.27, P \ 0.001, N = 978; Table 2) included the areas), and one physical environmental variable (slope).
effects of one demographic variable (rural population Rural population density, cattle, and slope affected nega-
density), two land use variables (cattle and protected tively the deforestation rate, indicating that forest loss
Fig. 2 Deforestation between
1990 and 2005 in the natural
regions of Colombia

123
Tropical deforestation in Colombia

Fig. 4 Deforestation rates per


municipality in Colombia for
the period 1990–2005

increased in municipalities with lower rural population The best model for deforestation in the Caribbean region
density, fewer cattle, and flatter slopes. In contrast, higher (R2 = 0.59, P \ 0.001, N = 204; Table 3) included the
deforestation rates were found where there was presence of effects of one demographic variable (urban population),
more protected areas in the municipality. one socioeconomic variable (unsatisfied basic needs), four
We carried out separate analyses of the drivers of land use variables (illicit crops, pastures, cattle, and fires),
deforestation for three natural regions. The best combina- and two physical environmental variables (slope and pre-
tion of explanatory variables for deforestation in Amazonia cipitation). Both the demographic and the socioeconomic
(R2 = 0.57, P \ 0.001, N = 55; Table 3) included the variables had positive effects on deforestation, indicating
effects of only two variables: one demographic variable that forest loss was higher in those municipalities with
(rural population density) and one concerning land use higher urban densities and more unsatisfied basic needs.
(fire). The former had a negative effect on the deforestation The presence of pastures and illicit crops also positively
rate, indicating that lower rural population densities were influenced deforestation rates, while cattle and fires had a
associated with higher deforestation rates. The latter significant negative effect on deforestation. Regarding the
showed that fires had a positive effect on deforestation physical environmental variables, the deforestation rate
rates. increased significantly with slope and precipitation,

123
Table 1 Summary of the variables considered in the deforestation analysis and data sources

123
Type Short name Units Spatial Description Source(s) Variable
resolution range

Demographic Urban population Number of Municipality Change in urban population density DANE 1985, 1993, 2005 -0.82–12.65
inhabitants/ha between 1985 and 2005
Rural density Number of Municipality Change in rural population density DANE 1985, 1993, 2005 -0.35–1.15
inhabitants/ha between 1985 and 2005
Socio economic Unsatisfied basic needs (NBI) % Municipality Percentage of population with unsatisfied DANE 2005 0.17–1
basic needs in 2005. In Latin America,
most countries consider minimum
household conditions, access to sanitary
services, access to primary education
and minimum household economic
capacity as basic needs
Land use Crops ha 30 m Total change in crop area between 1985 Calculated from IDEAM 2007 0–68,043
and 2005 derived from satellite image
classification
Pastures ha 30 m Total change in pasture area Calculated from IDEAM 2007 0–286,948
Illicit crops ha 10 m Area of coca (Erythroxylum coca) crops UNDOC (2006) through the 0–7,445
Colombian Integrated System for
Illicit Crops Monitoring project
Cattle Number Municipality Head of cattle per municipality in 2006 MADR 0–471,907
Fire Number Municipality Number of fire hotpots detected per FIRMS 0–13,748
municipality between 2000 and 2005
Mining kg Municipality Total gold and silver production in 2005 SIGOT-IGAC 2011 0–7,374,340
Protected area ha 1: 100.000 Area of each municipality under special IGAC 2005 Calculated 0–3,864,528
management as either national protected
area or indigenous reserve
Road density km/ha 1: 100.000 Density of roads in each municipality IGAC 2005 Calculated 0–8.36
Physical environment Slope Degrees 90 m Average maximum slope for each Calculated IGAC 2005 0.01–10.01
municipality
Water scarcity in dry years Unit less Municipality Index of water scarcity in a dry year IDEAM, 2000 1–5
Elevation *m 90 m Altitude above sea level IGAC 2005 0–3675
2
Precipitation mm 1 km Total annual precipitation Worldclim 2005 365–8,434
DANE National Administrative Department of Statistics, UNDOC United Nations Office on Drugs and Crime, MADR Agriculture and Rural Development Ministry, FIRMS Fire Information for
Resource Management System, SIGOT Geographic Information System for Planning and Territorial Development, IGAC Agustı́n Codazzi Geographical Institute, IDEAM Institute of
Meteorology and Environmental Studies, Worldclim global climate data
D. Armenteras et al.
Tropical deforestation in Colombia

Table 2 Statistics (F statistics used for testing the different variables deforestation rate (0.68 %) than Colombia (Urioste 2010).
and their significance P are given; the sign of the relationship between Most of the deforestation in Colombia occurred in transi-
each variable and the dependent variable is indicated in the last
tion zones between regions (Wassenaar et al. 2007;
column) of the general linear model for deforestation at the national
level as a function of variables representing demography (urban Armenteras et al. 2011; Rodrı́guez et al. 2011, 2012),
population and rural population), socioeconomy (unsatisfied basic reflecting intra-regional limits that tend to respond not to
needs), land use (crops, pastures, cattle, illicit crops, mining, fires, and political or administrative borders, but to geographical and
protected area), and the physical environment (road density, slope,
socioeconomic conditions. These deforestation hotspots
water scarcity in dry years, elevation, and precipitation)
tend to occur generally in lowland forests (except in the
F P Sign Caribbean region where forests remain only in highland
Intercept 3.6 0.057 - areas) with low population density, extreme poverty, low
ln urban population 0.3 0.588 ? income, and the presence of an illicit economy, a pattern
ln rural density 23.8 <0.001 - that is found elsewhere in the American tropics (Wassenaar
NBI 1.1 0.293878 ? et al. 2007).
ln crops 1.9 0.161314 ?
Our results suggest that, at the national level, defores-
ln pastures 2.9 0.084915 ?
tation is determined mainly by three drivers: human pop-
ulation density, land use, and slope. Lower rural densities,
Illicit crops 1.7 0.188 ?
fewer cattle, and lower slopes were associated with greater
ln cattle 53.6 <0.001 -
deforestation, suggesting that forest loss occurs to a greater
Fires 3.3 0.066 -
extent in remote areas with active colonization fronts and
Mining 0.9 0.344 ?
flatter topography. The results of this model are consistent
ln protected area 9.1 0.002 ?
with previous models in which deforestation is positively
Road density 0.8 0.370 ?
affected by impoverished populations without alternatives
Slope 11.6 <0.001 -
to clearing land as one of the early stages of the coloni-
Water scarcity 0.3 0.538072 ?
zation process (Rudel and Roper 1997). It is also in
Elevation 1.1 0.286 -
agreement with the fact that areas with lower and declining
Precipitation 0.9 0.336443 -
rural population and population density but with available
F and P values and the sign of the coefficient are shown agricultural land, do experience deforestation, as recently
Significant values at P \ 0.05 are in bold. n = 978 examined in a study of agricultural conversion for Latin
America (Carr et al. 2009). Further our results support the
importance of rural-to-urban migration in the region,
indicating that wetter and steeper municipalities had higher especially internal migrant colonists (Carr 2009). In gen-
forest loss. eral terms, deforestation in Colombia is occurring in flatter
The best combination of explanatory variables for areas, where cattle density and rural population density are
deforestation in the Orinoco region (R2 = 0.76, P \ 0.001, still low because settlements have not yet been established
N = 41; Table 3) included the effect of three land use (Rudel and Roper 1997) and also probably due to migration
variables (crops, illicit crops, mining) and one physical both to urban and frontier areas (Carr et al. 2009; Carr
environmental variable (elevation). All variables except 2009). Thus, Colombia as a whole is still at an early
mining had positive effects on the deforestation rate, developmental stage of natural resource extraction (the
indicating that the larger the area under cultivation (both exploitation phase).
legal and illegal) and the higher the elevation, the higher Despite the national trends described above, our results
the deforestation rate. In contrast, where more mining was confirm that the process of deforestation in Colombia is far
present, the deforestation rate was lower. from homogeneous and that spatial variability within the
country is high in terms of both the rate and the drivers of
deforestation. Deforestation rates in the years 1990–2005
Discussion were not uniform throughout the country. Variability in the
rate of forest loss is associated with the historical context of
Colombia had a higher deforestation rate (0.62 %) in the occupation and the geographical location of the regions,
period from 1990 to 2005 than the mean reported by FAO which influence the use of and the access to resources,
(2011) for South America for the years 1990–2010 respectively. We found that the main factors underlying
(0.45 %). Colombia’s annual net loss of forest is similar to deforestation clearly differ across different regions of
that reported for Bolivia (350,000 ha/year) and higher than Colombia (Table 3). The high heterogeneity of processes,
that reported for the neighboring country of Ecuador socio economic and demographic conditions, has also been
(74,300 ha/year), although this latter country has a higher found between countries in Latin America (Carr 2009).

123
D. Armenteras et al.

Table 3 Statistics (F statistics used for testing the different variables and their significance P are given; the sign of the relationship between each
variable and the dependent variable is indicated in the last column) of the general linear models for deforestation in three natural regions as a
function of variables representing demographics (urban population and rural population), socioeconomics (unsatisfied basic needs), land use
(crops, pastures, cattle, illicit crops, mining, fires, and protected area), and the physical environment (road density, slope, water scarcity in dry
years, elevation, and precipitation)
Region Amazonia Caribbean Orinoco
n = 55 n = 204 n = 41
Variable F P Sign F P Sign F P Sign

Intercept 1.2 0.270 - 41.9 \0.001 - 0.0 0.845 -


ln urban population 0.0 0.806 ? 8.9 0.003 ? 0.0 0.788 ?
ln rural density 4.8 0.033 - 2.4 0.119 - 0.2 0.641 -
NBI 0.2 0.627 - 17.2 <0.001 ? 0.0 0.855 -
ln crops 0.1 0.825 ? 0.6 0.439 - 4.7 0.038 ?
ln pastures 2.4 0.123 ? 9.6 0.002 ? 0.1 0.685 -
Illicit crops 2.1 0.154 ? 5.8 0.016 ? 12.9 0.001 ?
ln cattle 0.1 0.696 - 37.0 <0.001 - 1.9 0.180 -
Fires 4.9 0.031 ? 5.4 0.020 - 1.0 0.317 ?
Mining 0.5 0.473 - 0.0 0.844 - 5.1 0.031 -
ln protected area 0.1 0.771 ? 2.0 0.156 - 1.2 0.2732 ?
Road density 0.9 0.336 - 0.1 0.752 ? 0.7 0.396 ?
Slope 0.2 0.600 - 8.5 0.003 ? 3.2 0.085 -
Water scarcity 0.2 0.651 - 0.0 0.849 - 3.8 0.061 -
Elevation 0.4 0.485 ? 2.8 0.094 - 11.2 0.002 ?
Precipitation 0.01 0.907 - 16.9 <0.001 ? 1.8 0.187 -
F and P values and the sign of the coefficient are shown. The values of n for each region are as follows: Amazonia, 55; Caribbean, 204; Orinoco,
41. The model for the Pacific region was not computed given the low number of municipalities with data available in this region (n = 15)
Significant values at P \ 0.05 are in bold

Environmental and climatic conditions also vary across highest cattle productivity in Colombia (Viloria de la Hoz
regions within Colombia. In contrast to other regions such 2003). The establishment of commercial cattle ranching is
as the Andes where there is evidence of intensive agri- associated with the presence of fires that are used in tra-
cultural activities and forest logging since pre-Hispanic ditional management practices (Armenteras-Pascual et al.
(Etter et al. 2008; Armenteras et al. 2011), the process of 2011). The urban population in the Caribbean is high, with
immigration began recently (in the 1960s) in Amazonia, higher population densities within an inland area of
and this region contains currently active colonization and 100 km from the coast (217 inhabitants per km2, Roca and
deforestation fronts. The Pacific region has experienced Perez 2006). Both urban population size and unsatisfied
selective logging for decades but, together with the Ori- basic needs positively affect deforestation. Economic
noco region, has also been subjected to pressures from activities such as pastures and illicit crops also increase
biofuel projects over the last 10 years (Janssen and Rutz deforestation rates in the Caribbean region. Forest rem-
2011), which represents a set of newly emerging economic nants and thus deforestation hotspots are found in three
and development opportunities that might already have sub-regions (Sierra Nevada de Santa Marta, San Lucas, and
influenced deforestation in these regions, although popu- Perijá) where wetter and steeper slopes are found, whereas
lation density is low. Land use is described by as few as flatter areas tend to be occupied either by cattle ranching or
three factors in the Orinoco region and two factors in agricultural activities (Roca and Perez 2006). When land is
Amazonia either because other factors identified elsewhere scarce, highland areas in the Caribbean with steeper terrain
are not yet representative of the colonization process in provide farmers the only opportunity to clear forest and
these regions (e.g., cattle grazing, roads, and soy planta- expand agriculture in these marginal lands, and thus, rural
tions in the Amazon) or because data are not available populations tend to migrate to such areas. However, the
(e.g., logging for both Orinoco and Amazon). complex relationships between land use and climate in the
The Caribbean region is arguably the most transformed Caribbean region remain poorly understood.
region, with a long history of occupation and land use The Amazonian and Orinoco regions are both
traditionally driven by cattle grazing. This region has the scarcely populated areas with small-scale agriculture, cattle

123
Tropical deforestation in Colombia

ranching, and active agricultural frontiers (including illegal different regional amounts and rates of deforestation result
crops) in many municipalities (Barbosa and Fearnside from current different drivers at the municipality level, and
2005; Armenteras et al. 2006; Romero-Ruiz et al. 2009). thus, the implications for land use and conservation plan-
Oil palm expansion is starting to affect these areas. Indeed, ning can be addressed at this geopolitical scale.
deforestation in the Orinoco region is strongly affected by The overall conclusion of this study is that the factors
legal and illegal crops, which impact the integrity of gal- underlying deforestation vary across Colombia as a
lery forests, and also by some mining activities that could reflection of environmental differences and also of differ-
impact montane forests in the higher parts of the basin. Our ent stages of development, which are similar to the changes
results for Amazonia suggest that within this region, over time that have been reported more broadly for Latin
deforestation occurs primarily in remote areas where set- America (Rudel et al. 2009). Thus, deforestation must be
tlements have not yet been established and a floating col- analyzed bearing in mind intra-national variation, particu-
onist population moves around, creating new colonization larly differences in the relative importance of multiple
hotspots mainly using river networks (Armenteras et al. factors from one region to another, to avoid misdirecting
2006; Etter et al. 2006). Fire remains a driver of defores- national and regional policies designed to reduce defores-
tation, as has been reported elsewhere in the Amazonian tation and forest degradation. Adding to the complexity of
region (Cochrane and Barber 2009; Aragão et al. 2008) and this issue, Colombia is currently promoting different levels
also in the Colombian Amazon (Armenteras and Retana of development in agriculture, mining, energy, road infra-
2012) where fire is an indicator of slash and burn activities structure, and technology, prioritizing each of these areas
associated with deforestation and with early stages of differently in the different regions. There is also a need to
natural resource exploitation as found for other areas of the consider the fact that many of the variables used in this
Amazon where pastures and small-scale agricultural study are dynamic over time and themselves are driven by
activities (Müller et al. 2011; Egler et al. 2012; Armenteras other factors, indirect drivers such as global economic
et al. 2013). conditions that will affect the way cattle and crops respond.
Despite the fact that in the national-level analysis, illicit Even including global market forces, much uncertainty
crops did not appear to have an effect on deforestation remains as to whether forest degradation and deforestation
(Table 2), they were found to be an important factor at the must be addressed nationally or regionally, or whether
regional level in two of three regions (Table 3), in agree- strategies to mitigate these problems should instead be
ment with recent reports (Dávalos et al. 2011). Surpris- focused on active colonization fronts. The latter approach
ingly, this factor did not appear to be important in the might be appropriate if colonists are given the opportunity
Amazonian region despite the extensive area under culti- to access alternative income sources through productive
vation in this region. One possible explanation is that the lands and wealth creation in the private sector (Killeen
effect of illicit crops in Amazonia was not captured by our et al. 2011), reducing migration toward forest areas. Thus,
model or either they have smaller impact on total defor- land tenure and access might be the key. Because global-
estation than previously expected (see, for a similar con- ization is introducing new drivers of deforestation (Rudel
clusion, Armenteras et al. 2013). Alternatively, there might et al. 2009; Killeen et al. 2011), such as biofuel production,
be a link between rural population density, deforestation, mining, and hydrocarbon extraction in Colombia, the
and an illegal economy such that, as Dávalos et al. (2011) extent of the impact that these factors will have on forests
suggest, coca growing might be a consequence (attractor) and how the country should cope with them both nationally
of poverty and not a cause of deforestation. Nevertheless, and regionally remain unknown.
the impact of the expansion of illegal crops on forest
fragmentation and deforestation has been proved in the past Acknowledgments We thank the Instituto de Hidrologı́a, Meteo-
rologı́a y Estudios Ambientales (IDEAM) for providing the defor-
in many areas of Latin America (Armenteras et al. 2006, estation data maps from the national Colombia REDD (Reducing
2009; Bradley and Millington 2008; Rodrı́guez et al. 2011). Emissions from Deforestation and Degradation) project funded by
Our results corroborate the conclusions of Aguiar et al. the Gordon and Betty Moore Foundation and its team members
(2007) for the Amazonian region, highlighting the high Maria Claudia Garcı́a, Maria Fernanda Ordoñez, Gustavo Galindo,
and Diana Vargas. We also thank the MODIS team for providing
regional heterogeneity in deforestation. Within Colombia, continuous access to active fire datasets. Thanks to Carol Franco for
a previous study based on static forest cover suggested that her continuous support in data handling. We also want to thank you
important regional differences exist in past land use, but the anonymous reviewers for their valuable comments and sugges-
suggested that within regions, deforestation is driven tions to improve the quality of the paper. The research was funded
by the Universidad Nacional de Colombia through the postdoctoral
mostly by common biophysical variables, distances to position of Nelly Rodrı́guez and a grant from the CONVOCATO-
towns, and the presence of roads, reflecting past clearing RIA NACIONAL PARA EL FORTALECIMIENTO DE ALIAN-
events spanning several 100 years (Etter et al. 2006). The ZAS ESTRATÉGICAS INTERINSTITUCIONALES. UN - 2012,
results of our study add to our understanding of how Project 12.

123
D. Armenteras et al.

References tropical biodiversity hotspots. Environ Sci Technol 45:


1219–1227
Achard F, Eva H, Stibig HJ, Mayaux P, Gallego J, Richards T, Davies DK, Ilavajhala S, Wong MM, Justice CO (2009) Fire
Malingreau JP (2002) Determination of deforestation rates of the information for resource management system: archiving and
world’s humid tropical forests. Science 297:999–1002 distributing MODIS active fire data. IEEE T Geosci Remote
Aguiar AP, Camara G, Sobral MI (2007) Spatial statistical analysis of 47:72–79
land-use determinants in the Brazilian Amazonia: exploring Defries RS, Rudel T, Uriarte M, Hansen M (2010) Deforestation
intra-regional heterogeneity. Ecol Model 209:169–188 driven by urban population growth and agricultural trade in the
Angulo CP (2011) Dinámica de la deforestación en Venezuela: twenty-first century. Nature Geosci 3:178–181
análisis de los cambios a partir de mapas históricos. Interciencia Egler M, Egler CAG, Franz B, Muylaert de Araujo MS, Freitas MAV
36:578–587 (2012) Indicators of deforestation in the Southern Brazilian Pre-
Aragão LEOC, Malhi Y, Barbier N, Lima A, Shimabukuro Y, Amazon. Reg Environ Change. doi:10.1007/s10113-012-0331-4
Anderson L, Saatchi S (2008) Interactions between rainfall, Etter A, McAlpine C, Wilson L, Phinn S, Possingham H (2006)
deforestation and fires during recent years in the Brazilian Regional patterns of agricultural land use and deforestation in
Amazonia. Philos T R Soc B 363:1779–1785 Colombia. Agric Ecosyst Environ 114:369–386
Armenteras D, Rudas G, Rodrı́guez N, Sua S, Romero M (2006) Etter A, McAlpine C, Possingham H (2008) Historical patterns and
Patterns and causes of deforestation in the Colombian Amazon. drivers of landscape change in Colombia since 1500: a
Ecol Indic 6:353–368 regionalized spatial approach. Ann Assoc Am Geogr 98:2–23
Armenteras D, Rodrı́guez N, Retana J (2009) Are conservation FAO (2011) State of the world’s forests. FAO, Rome. http://www.
strategies effective in avoiding the deforestation of the Colom- fao.org/docrep/013/i2000e/i2000e.pdf
bian Guyana Shield? Biol Conserv 142:1411–1419 Fearnside PM (1984) Initial soil quality conditions on the Trans
Armenteras D, Rodrı́guez N, Retana J, Morales M (2011) Under- Amazon Highway of Brazil and their simulation in models for
standing deforestation in montane and lowland areas of the estimating human carrying capacity. Tropical Ecol 25:1–21
Andes in Colombia. Reg Environ Change 11:693–705 Fearnside PM (1993) Deforestation in Brazilian Amazonia: the effect
Armenteras-Pascual D, Retana J, Molowny-Horas R, Roman-Cuesta of population and land tenure. Ambio 22:537–545
RM, Gonzalez-Alonso F, Morales-Rivas M (2011) Character- Forrest JL, Sanderson EW, Wallace R, Lazzo TMS, Cerveró LHG,
ization of fire spatial patterns interactions with climate and Coppolillo P (2008) Patterns of land cover change in and around
vegetation in Colombia. Agr For Meteorol 151:279–289 Madidi National Park, Bolivia. Biotropica 40:285–294
Armenteras D, Retana J (2012) Dynamics, patterns and causes of fires Geist HJ, Lambin EF (2001) What drives tropical deforestation? A
in Northwestern Amazonia. PLoS One 7:e35288 meta-analysis of proximate and underlying causes of deforesta-
Armenteras D, Rodrı́guez N, Retana J (2013) Landscape dynamics in tion based on subnational case study evidence. LUCC Interna-
northwestern amazonia: an assessment of pastures, fire and illicit tional Project Office, LUCC Report Series no. 4, Louvain-la-
crops as drivers of tropical deforestation. PLoS One 8(1):e54310. Neuve, Belgium. http://www.pik-potsdam.de/members/cramer/
doi:10.1371/journal.pone.0054310 teaching/0607/Geist_2001_LUCC_Report.pdf
Barbosa RI, Fearnside PM (2005) Fire frequency and area burned in Geist HJ, Lambin EF (2002) Proximate causes and underlying driving
the Roraima savannas of Brazilian Amazonia. For Ecol Manag forces of tropical deforestation. Bioscience 52:143–150
204:371–384 Grau HR, Aide M (2008) Globalization and land-use transitions in
Bradley AV, Millington AC (2008) Coca and colonists: quantifying Latin America. Ecol Soc 13:16
and explaining forest clearance under coca and anti-narcotics Halekoh U, Højsgaard S (2004) Generalized linear models. Master of
policy regimes. Ecol Soc 13:31 Applied Statistics, Tjele, Denmark
Butler RA, Laurance WF (2008) New strategies for conserving Hansen MC, Stehman SV, Potapov PV (2010) Quantification of
tropical forests. Trends Ecol Evol 23:469–472 global gross forest cover loss. PNAS 107:8650–8655
Calvo-Alvarado J, McLennan B, Sánchez-Azofeifa A, Garvin T Herrador Valencia D, Boada i Juncà M, Varga Linde D, Mendizábal
(2009) Deforestation and forest restoration in Guanacaste, Costa Riera E (2011) Tropical forest recovery and socio-economic
Rica: putting conservation policies in context. For Ecol Manag change in El Salvador: an opportunity for the introduction of
258:931–940 new approaches to biodiversity protection. Appl Geogr 31:
Carr D (2009) Rural migration: the driving force behind tropical 259–268
deforestation on the settlement frontier. Prog Hum Geo 33: Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very
1–31 high resolution interpolated climate surfaces for global land
Carr DL, Suter L, Barbieri A (2005) Population dynamics and tropical areas. Int J Climatol 25:1965–1978
deforestation: state of the debate and conceptual challenges. Huang C, Kim S, Altstatt A, Townshend JRG, Davis P, Song K,
Popul Environ 27:89–113 Tucker CJ et al (2007) Rapid loss of Paraguay’s Atlantic forest
Carr DL, Lopez AC, Bilsborrow RE (2009) The population, and the status of protected areas A Landsat assessment. Remote
agriculture, and environment nexus in Latin America: country- Sens Environ 106(4):460–466
level evidence from the latter half of the twentieth century. Popul Huang C, Kim S, Song K, Townshend JRG, Davis P, Altstatt A,
Environ 30:222–246 Rodas O et al (2009) Assessment of Paraguay’s forest cover
Cochrane M, Barber CP (2009) Climate change, human land use and change using Landsat observations. Global Planet Change
future fires in the Amazon. Glob Change Biol 15:601–612 67(1–2):1–12
Cotte Poveda A, Pardo Martı́nez CI (2011) Trends in economic IDEAM, IGAC, IAvH, Invemar, Sinchi, IIAP (2007) Ecosistemas
growth, poverty and energy in Colombia: long-run and short-run continentales, costeros y marinos de Colombia. Instituto de
effects. Energy Sys 2:281–298 Hidrologı́a, Meteorologı́a y Estudios Ambientales, Instituto
Chomitz KM, Thomas TS (2003) Determinants of land use in Amazonia: Geográfico Agustı́n Codazzi, Instituto de Investigación de
a fine scale spatial analysis. Am J Agr Econ 85:1016–1028 Recursos Biológicos Alexander von Humboldt, Instituto de
Dávalos LM, Bejarano AC, Hall MA, Correa HL, Corthals AP, Investigaciones Marinas y Costeras José Benito Vives de
Espejo OJ (2011) Forests and drugs: coca-driven deforestation in Andréis, Instituto Amazónico de Investigaciones Cientı́ficas

123
Tropical deforestation in Colombia

Sinchi e Instituto de Investigaciones Ambientales del Pacı́fico Roca AM, Perez GJ (2006) Geografı́a fı́sica y poblamiento en la
Jhon von Neumann. Bogotá, DC Costa Caribe colombiana. Documentos de trabajo sobre Eco-
Instituto de Hidrologı́a, Meteorologı́a y Estudios Ambientales nomı́a Regional No 73, Junio 2006. Banco de la República,
[IDEAM] (2011) Memoria Técnica Cuantificación de la Defor- Colombia
estación histórica para Colombia a nivel nacional, escalas gruesa Rodrı́guez N, Armenteras D, Retana J, Molowny-Horas R (2011)
y fina. Bogotá, p 85 (in press) Patterns and trends of forest loss in the Colombian Guyana.
Instituto Geográfico Agustı́n Codazzi [IGAC] (2005) 1:500.00 Biotropica 44(1):123–132
Official cartography Rodrı́guez N, Armenteras D, Retana J (2012) Land use and land cover
Instituto Geográfico Agustı́n Codazzi [IGAC] (2011) Sistema de change in the Colombian Andes: dynamics and future scenarios.
Información Geográfica para la Planeación y el Ordenamiento J Land Use Sci. doi:10.1080/1747423X.2011.650228
Territorial Nacional SIGOT, Bogotá, Colombia Romero-Ruiz M, Etter A, Sarmiento A, Tansey K (2009) Spatial and
Janssen R, Rutz DD (2011) Sustainability of biofuels in Latin temporal variability of fires in relation to ecosystems, land tenure
America: risks and opportunities. Energ Policy. doi:10.1016/j. and rainfall in savannas of northern South America. Glob
enpol.2011.01.047 Change Biol 16:2013–2023
Jarvis A, Touval J, Schmitz MC, Sotomayor L, Hyman GG (2010) Rudel TK (2006) Shrinking tropical forests, human agents of change
Assessment of threats to ecosystems in South America. J Nat and conservation policy. Conserv Biol 20:1604–1609
Conserv 18:180–188 Rudel TK (2007) Changing agents of deforestation: from state-
Killeen TJ, Schroth G, Turner W, Harvey CA, Steininger MA (2011) initiated to enterprise driven processes, 1970–2000. Land Use
Stabilizing the agricultural frontier: leveraging REDD with Policy 24:35–41
biofuels for sustainable development. Biomass Bioenerg 35(12): Rudel TK, Roper J (1997) The paths to rain forest destruction: cross-
4815–4823 national Patterns of tropical deforestation, 1975–90. World Dev
Kindermann G, Obersteiner M, Sohngen B, Sathaye J, Andrasko K, 25:53–65
Rametsteiner E, Schlamadinger B, Wunder S, Beach R (2008) Rudel TK, Defries R, Asner GP, Laurance WF (2009) Changing
Global cost estimates of reducing carbon emissions through drivers of deforestation and new opportunities for conservation.
avoided deforestation. PNAS 105:10302–10307 Conserv Biol 23(6):1396–1405
Laurance WF, Albernaz AKM, Da Costa C (2001) Is deforestation Statsoft (2001) STATISTICA 6.0 package. Statsoft. Tulsa, USA
accelerating in the Brazilian Amazon? Environ Conserv 28: Steininger MK, Tucker CJ, Townshend JRG, Killeen TJ, Desch A,
305–311 Bell V, Ersts P (2001) Tropical deforestation in the Bolivian
Leica Geosystems (2005) ERDAS Imagine 9.1. Leica Geosystems, Amazon. Environ Conserv 28:127–134
GIS and mapping division, Atlanta, GA Stevens K, Campbell L, Urquhart G, Kramer D, Qi J (2011)
López S, Sierra R (2010) Tropical deforestation in the Ecuadorian Examining complexities of forest cover change during armed
Chocó: logging practices and socio-spatial relationships. Geo- conflict on Nicaragua’s Atlantic Coast. Biodivers Conserv
grap Bull 51:3–22 20:2597–2613
Mittermeier RR, Myers N, Thomsen JJB, da Fonseca GAB, Olivieri S United Nations Office on Drugs and Crime—UNODC (2009)
(1998) Biodiversity hotspots and major tropical wilderness areas: Colombia: Monitoreo de Cultivos de Coca 2009. UNODC,
approaches to setting conservation priorities. Conserv Biol Bogotá
12:516–520 Urioste A (2010) Deforestación en Bolivia. Una amenaza mayor al
Monzón-Alvarado C, Cortina-Villar S, Schmook B, Flamenco- cambio climático. Fundación Friedrich Ebert. Working paper.
Sandoval A, Christman Z, Arriola L (2012) Land-use decision- http://library.fes.de/pdf-files/bueros/bolivien/07570.pdf
making after large-scale forest fires: analyzing fires as a driver of Viloria de la Hoz J (2003) La ganaderı́a bovina en las Llanuras del
deforestation in Laguna del Tigre National Park, Guatemala. Caribe colombiano. Documentos de trabajo sobre economı́a
Appl Geogr 35:43–52 regional. Banco de la República, No. 40, Cartagena de Indias,
Müller R, Müller D, Schierhorn F, Gerold G, Pacheco P (2011) Colombia
Proximate causes of deforestation in the Bolivian lowlands: an Wassenaar T, Gerber P, Verburg PH, Rosales M, Ibrahim M,
analysis of spatial dynamics. Reg Environ Change 2:445–459 Steinfeld H (2007) Projecting land use changes in the neotropics:
Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GA, Kent J the geography of pasture expansion into forest. Global Environ
(2000) Biodiversity hotspots for conservation priorities. Nature Chang 17:86–104
403:853–858 Wright S, Samaniego M (2008) Historical, demographic, and
Portillo-Quintero C, Sanchez M, Valbuena C, Gonzalez YY, Larreal economic correlates of land-use change in the Republic of
JT (2012) Forest cover and deforestation patterns in the Northern Panama. Ecol Soc 13(2):17
Andes (Lake Maracaibo Basin): a synoptic assessment using Zeledon EB, Kelly NM (2009) Understanding large-scale deforesta-
MODIS and Landsat imagery. Appl Geog 35:152–163 tion in southern Jinotega, Nicaragua from 1978 to 1999 through
Redo D, Joby Bass JO, Millington AC (2009) Forest dynamics and the examination of changes in land use and land cover. J Environ
the importance of place in western Honduras. Appl Geog Manag 90:2866–2872
29:91–110

123

You might also like