Global Environmental Change 46 (2017) 139–147

Contents lists available at ScienceDirect

Global Environmental Change

journal homepage: [Link]/locate/gloenvcha

Deforestation dynamics and drivers in different forest types in Latin MARK

America: Three decades of studies (1980–2010)
⁎
Dolors Armenterasa, , Josep María Espeltab, Nelly Rodrígueza, Javier Retanab,c
a
Grupo de Ecología del Paisaje y Modelación de Ecosistemas ECOLMOD, Departamento de Biología, Facultad de Ciencias, Universidad Nacional de Colombia, Sede
Bogotá, Colombia
b
CREAF-Centre for Ecological Research and Forestry Applications, 08193 Cerdanyola del Valles, Barcelona,Spain
c
Unitat d'Ecología, Universitat Autonòma de Barcelona, 08193 Cerdanyola del Valles, Barcelona, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: Over the last decades there have been a considerable number of deforestation studies in Latin America reporting
Forest loss lower rates compared with other regions; although these studies are either regional or local and do not allow the
Deforestation rates comparison of the intraregional variability present among countries or forest types. Here, we present the results
Tropical deforestation obtained from a systematic review of 369 articles (published from 1990 to 2014) about deforestation rates for 17
Factors
countries and forest types (tropical lowland, tropical montane, tropical and subtropical dry, subtropical tem-
Causes
perate and mixed, and Atlantic forests). Drivers identified as direct or indirect causes of deforestation in the
literature were also analysed. With an overall annual deforestation rate of −1.14 ( ± 0.092 SE) in the region, we
compared the rates per forest type and country. The results indicate that there is a high variability of forest loss
rates among countries and forest types. In general, Chile and Argentina presented the highest deforestation rates
(−3.28 and −2.31 yearly average, respectively), followed by Ecuador and Paraguay (−2.19 and −1.89 yearly
average, respectively). Atlantic forests (−1.62) and tropical montane forests (−1.55) presented the highest
deforestation rates for the region. In particular, tropical lowland forests in Ecuador (−2.42) and tropical dry
forests in Mexico (−2.88) and Argentina (−2.20) were the most affected. In most countries, the access to
markets and agricultural and forest activities are the main causes of deforestation; however, the causes vary
according to the forest types. Deforestation measurements focused at different scales and on different forest types
will help governments to improve their reports for international initiatives, such as reducing emissions from
deforestation and forest degradation (REDD+) but, more importantly, for developing local policies for the
sustainable management of forests and for reducing the deforestation in Latin America.

1. Introduction area between 2000 and 2005 of 101.1 million hectares. One of the latest
published global figures of deforestation indicates global net losses of
The destruction of tropical forests has received worldwide attention tropical forests of 6.1 million hectares per year for the 1990–2000
because of the well-known, unique role they play in ecological terms, period (0.377%) and 5.9 million hectares per year during the 2000 s
the diversity of functions they provide and, above all, the continuing (0.384% annually; Achard et al., 2014)
threat to its existence, which directly affects the net carbon emissions At the continental level, (Achard et al., 2002) reported a defor-
derived from deforestation and degradation (Houghton, 2012). In ad- estation rate of 0.38% for Latin America, 0.43% for Africa, 0.91% for
dition, the deforestation rates are far from being uniform across the Southeast Asia and an overall rate of 0.52%. Brazil and Indonesia ac-
world and depend on the different analyses and sources of data used for counted for 20.3% of the loss of tropical forests in 1980, 25.7% during
their calculation. The Food and Agriculture Organization of the United the 1990s and the 40.7% between 2000 and 2005. Since the year 2000,
Nations (FAO, 2011) estimated a net global deforestation of 0.20% in several reports suggest that these tropical regions have significantly
the decade from 1990 to 2000, 0.12% between 2000 and 2005 and reduced their deforestation rate (Achard et al., 2014; Food and Agri-
0.14% between 2005 and 2010, with a net loss of 5.2 million hectares culture Organization of the United Nations (FAO, 2010a). In the 1990s,
from the year 2000 to 2010. Instead, (Hansen et al., 2010) indicated a several authors estimated a decrease in the rate of deforestation and an
rate of 0.6% of annual forest loss and an estimated loss of global forest increase in the forest area through planting or natural expansion and

⁎
Corresponding author at: Departamento de Biología, Edificio 421, Oficina 223, Universidad Nacional de Colombia, 111321 Bogotá, Colombia.
E-mail address: darmenterasp@[Link] (D. Armenteras).

[Link]
Received 10 November 2016; Received in revised form 5 June 2017; Accepted 1 September 2017
0959-3780/ © 2017 Elsevier Ltd. All rights reserved.
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

recovery of existing forests in Asia, Africa and Latin America (FAO, international markets (increased consumption of corn, sugar cane, palm
2010b; Rudel et al., 2009). In Latin America, the largest reported net oil and biofuels) and urbanisation associated with urban population
loss of forests in recent decades occurred from 2000 to 2010 (4 million growth, also play a crucial role (Ramankutty et al., 2007; Rudel et al.,
ha/year), with an increase from 2000 to 2005 (FAO, 2010). A more 2009). Other drivers of change in the tropics are associated with gen-
recent global study (Achard et al., 2014) indicates that the annual net eral patterns of land use and conversion of forest to pasture (Ra-
deforestation for humid and dry forests in Central and South America mankutty et al., 2006). In Latin America, geographic, socio-economic
between 2000 and 2010 was 1.92 and 0.92 million of hectares, re- factors and biophysical parameters have been proposed as the most
spectively. These values are higher than those from the previous decade important factors of recent changes in the land use (Wassenaar et al.,
in the case of humid forests (1.86 million of hectares in 1990–2000) and 2007). To a lesser extent (and impact), other factors have been pro-
lower than those from dry forests (0.99 million of hectares). The lack of posed, such as accessibility, demand for domestic and international
comparable historical national forest inventory data has made these markets, growth in population density, particularly in lowland transi-
global datasets one of the few available datasets to compare across tional areas where the most active deforestation frontiers are found
nations. Yet, the different approaches constitute a challenge for com- (Armenteras et al., 2011, 2006; Gomez-Peralta et al., 2008; Rudel et al.,
paring amongst them. Furthermore, most existing studies report the 2009; Wassenaar et al., 2007).
deforestation in different ways (e.g., total deforestation annual rate, The majority of recent regional or cross national studies about tro-
total rate, gross, net loss in hectares, among others), which makes the pical deforestation have either focused on the general forest/non forest
comparison of the dynamics of forest loss complex and, a priori, less pattern or on a maximum of two forest types (Achard et al., 2014; Aide
direct. A solution to this issue is to use the same standardised defor- et al., 2013; Hansen et al., 2013, 2008), e.g., the comparison of woody
estation rate, such as the one proposed by Puyravaud (2003). vegetation and plantations vs. mixed-woody vegetation (Aide et al.,
Addressing the causes of tropical deforestation requires not only 2013) or humid vs. dry tropical forests (Achard et al., 2002, 2014).
reliable figures on deforestation rates but also an understanding of the However, it is well known that different forest types greatly vary from
socioeconomic dynamics of the regional and local scales. The reported region to region together with the geographical (latitude, altitude,
causes and agents of forest loss act usually at different scales, and the biotic components, microclimate, among others) and socioeconomic
governments are challenged by the clear limitations that exist in terms factors that affect them. Because of this high variability of forest types,
of homogeneous information, from both the social and environmental environmental conditions and human dimensions, it seems reasonable
points of view. Furthermore, tropical forests play a vital role in bal- to undertake an analysis that would help to compile the deforestation
ancing the global greenhouse gas (GHG) emissions due to their ample rates and untangle the different dynamics and causes across the dif-
above- and below-ground carbon storage (Houghton, 2005). For a long ferent tropical forests types in Latin America. The aim of this work is to
time forests and climate change were dealt with on separate interna- (i) analyse deforestation rates in the different forest types and countries,
tional policy tracks, on the one hand on several of the UN Tropical (ii) explore the impact of the Kyoto protocol on deforestation rates and,
Forestry Action Programmes and national forest programmes, and on finally, (iii) evaluate the drivers of deforestation per forest type and
the other hand on the UN Framework Convention on Climate Change country. Based on the previous information, our hypotheses for these
(UNFCCC) (Buizer et al., 2014). The Kyoto Protocol recognised the objectives are: (i) dry forests have higher deforestation rates than more
importance of forests in climate change mitigation and somehow re- humid and lowland ones, (ii) deforestation rates have decreased after
defined international climate and forest politics that eventually led into the Kyoto protocol, and (iii) the expansion of agriculture and pastures
current mechanisms REDD+ (Buizer et al., 2014). In the context of remains as the major driver of forest loss across Latin America. We have
national policies of the countries that signed the Kyoto Protocol (1997), conducted a systematic and comprehensive review of peer-reviewed
the governments have committed to reduce the drivers of forest change publications related to deforestation studies in Latin America for the
(Kissinger et al., 2012). For the success of current international pro- last 30 yrs (1980–2010). We have compared the deforestation rates and
grammes that aim at reducing emissions from deforestation and to in- their drivers among different countries and forest types using a stan-
crease the atmospheric GHG removal by forests (e.g., reducing emis- dardized formula to calculate the annual rate of change of forest cover
sions from deforestation and forest degradation −REDD+), a better (Puyravaud, 2003). For a sub-set of countries, we have also conducted a
understanding of the causes of forest loss must be observed as potential temporal comparison of deforestation rates taking into account the year
opportunities to promote forest conservation and climate change miti- of one of the most relevant political agreements concerning environ-
gation. mental policies (i.e., pre- or post-Kyoto Protocol).
Several advances have been undertaken regarding the explanation
of the causes of deforestation patterns in the tropics (Geist and Lambin, 1.1. Methodology
2001). The agricultural expansion in forest frontiers is probably the
most cited cause in the literature as the main direct factor of forest loss 1.1.1. Study area
in the world (Gibbs et al., 2010), followed by other factors such as the The study area includes most tropical, subtropical and temperate
conversion of forest to pastures, logging for obtaining energy sources forests of Latin America, from Mexico in the north to Argentina and
and construction or expansion of infrastructures (Carr, 2004). Behind Chile in the south (see the list of countries included in Table 1). Un-
these direct causes, the understanding of causes and agents of defor- fortunately, small countries such as Suriname, Belize or Guyana were
estation has evolved to include more distant or underlying drivers of not included in the analysis due to their limited availability of in-
deforestation that involve economic, demographic, technological, cul- formation.
tural and political factors that operate at multiple scales and that differ
among regions (Geist and Lambin, 2001; Mather et al., 1999; Meyfroidt 1.1.2. Data collection
et al., 2013). Rudel et al. (2009) identified that between 1960 and 1980 A comprehensive search was conducted in three databases: a)
the forces behind the deforestation were social. The results of this study Scopus, b) Web of Science, and c) Google Scholar. To minimize bias we
showed an increase in deforestation rates, especially in areas where explicitly stated the hypotheses and the methodological approaches
colonisation schemes promoted the construction of roads and new prior to undertaking the research without prior knowledge of the data
settlements for rural populations. Hence, it is necessary to recognise the (Silagy et al., 2002). The first search criterion was the year of pub-
deforestation as a dynamic process associated with social, political and lication. We focused the search from 1990 to 2014, considering that
economic changes. These factors change over time, and the trends of even studies published during these years could contain forest cover
forest loss from the 1990s to the present reflect the changes in these dates prior to the date of publication. As second criteria, the following
causes. More recent processes, such as globalisation, the demand for keywords were considered: REDD, deforestation, deforestation drivers,

140
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

Table 1 considered due to the local and narrow scope for our analysis. When the
Mean ( ± SE) deforestation rate per country. N = number of case studies. drivers of deforestation were also reported or discussed, we also in-
cluded them in the database. If an article stated several deforestation
Country Mean deforestation rate ( ± SE) n
cases, years, types of forests or study areas, all that information was
Argentina −2.31 ± 0.41 47 considered as separate cases. For the purpose of standardisation and
Bolivia −0.01 ± 0.20 201 comparison of statistics, the forest area data and time were used to
Brazil −1.22 ± 0.29 102
recalculate a standardised deforestation rate (r) using the following
Chile −3.28 ± 0.49 33
Colombia −1.70 ± 0.67 18 formula (Puyravaud, 2003):
Costa Rica 0.30 ± 0.59 23
Ecuador −2.19 ± 0.29 97 1 A
Deforestation rate(r) = ⋅Ln ⎛ 2 ⎞⋅100
⎜ ⎟

Guatemala −0.49 ± 0.34 70 (t2 − t1) ⎝ A1 ⎠

Honduras −1.18 ± 0.51 30
Mexico −1.34 ± 0.17 286
where A1 and A2 are the forest areas (in has.) in the years t1 and t2,
Nicaragua −0.15 ± 0.81 12
Panama −0.40 ± 0.64 19
respectively (e.g., for a period from t1 = 1985 to t2 = 1990, A1 and A2
Paraguay −1.89 ± 0.36 62 are the values of forest cover in 1985 and 1990, respectively). This
Peru −0.15 ± 0.72 15 standardisation of deforestation rates makes the annual changes com-
Venezuela −0.36 ± 0.51 30 parable across countries (Fig. 2), forest types and periods. In any case,
the rates should not be confused with the total deforested area, which is
not assessed in this work given the type of data analysed.
afforestation, land use, proximate causes, causes deforestation and
We were interested in analysing the potential changes in defor-
forest loss. These keywords were optional search criteria (OR) and non-
estation rates of the region according to the major changes in world
compulsory contrary to the country name that was compulsory (AND)
environmental policies. Therefore, for every record in the dataset, we
in each case. In Scopus, the search fields were Article, Title, Abstract
calculated the number of years during which the study had been con-
and Keyword. In Web of Science, the search fields were Title and Topic.
ducted before or after the Kyoto Protocol Summit (1997) because we
In Goggle Scholar, the advanced search words were entered in the field
consider that protocol as a relevant inflection point in environmental
“with at least one of the words” and the search field was the “All items”.
policies. Negative values indicate that the study had ended before the
Papers in Spanish or Portuguese with keywords or abstract in English
Kyoto Protocol (e.g., a study ended in 1995 had a −2 score), while
were also included. With the criteria discussed above and after re-
positive values indicate that the study had begun after the summit (e.g.,
moving duplicates, 369 articles were compiled as potential sources of
a study that started in 1998 and ended in 2000 had a + 3 score). For
information.
those studies expanding both before and after 1997, we calculated the
difference in the number of years; this number was the score retained
1.1.3. Definition of the different forest types and used in our analysis for choosing the period before or after Kyoto
Several papers alluded to different types of forests. Given the dif- (e.g., for a study initiated in 1990 that ended in 2000, the score was −7
ferent classifications of forests that exist in the region, in general, and in +3 = −4, i.e., this period before Kyoto was assigned)
each country, in particular, we standardised these forest types into 6
broad types adding both latitudinal and altitudinal criteria (Armenteras
et al., 2003; Eva et al., 2004; Holdridge, 1967; Olson et al., 2001; 1.1.5. Causes of deforestation
Vreugdenhil et al., 2002): From the 186 papers with information on deforestation rates, 12 did
not mention any causes or drivers and were excluded from this part of
1. Tropical lowland forests (LW): forests located below 1000 m (such the analysis. We assessed each one of the 174 papers and extracted the
as tropical forests, humid rainforests, tropical lowland forests, tro- information on drivers in a matrix where the causes of deforestation
pical moist forests and floodplain forests). were grouped into direct and indirect drivers following Geist and
2. Tropical montane forests (TMF): forests in highlands above 1000 m Lambin (2002). The direct drivers were divided into geophysical vari-
(such as tropical moist montane forests, montane cloud forests, ables, infrastructures, agricultural expansion, cattle grazing, forestry,
upland forests and cloud forests) with over 2000 mm of rain per aquaculture, natural disasters, fire and mining. The indirect drivers
year. were grouped into demographic (population pressure), economic
3. Tropical and subtropical dry forests (TDF): forests in tropical and (markets and prices) and policy factors (development and agricultural
subtropical regions with 250–2000 mm of rain per year and a strong policies). When we could not associate a driver to one of the previous
dry season of at least 3–4 months (such as dry forests and seasonal categories, we assigned it to the “Others” category.
or wet-dry tropical forest).
4. Subtropical temperate and mixed forests (TF): forests in the sub-
1.1.6. Data analysis
tropics dominated by coniferous, temperate and mixed forests.
The literature review allowed us to have an initial data set of 1174
5. Atlantic forest (AF): a set of formations such as deciduous forests,
standardised deforestation rates. Differences in deforestation rates
ombrophilous dense forests, mixed ombrophilous to drier cerrado-
among forest types and among countries were analysed by means of a
type forest cover.
general linear model (GLM) including the geographical extent (surface)
6. Mangrove forests (MF): tropical and subtropical coastal forests
covered by each study as a covariate. In addition, we also run separate
where freshwater meets saltwater.
GLMs for certain sub-sets of data corresponding to two different forest
types (TDF and LW), which allowed us to include at least four different
1.1.4. Computation of deforestation rates countries with at least 15 cases. In these analyses we included the fol-
From the literature search, we systematically analysed each of the lowing variables: country, surface covered by the study, number of
publications, and only those reporting the initial extent of forest and the years during which the study was carried out, and the period when the
loss/gain of area for a certain period of time were selected. After this study was conducted (i.e., pre- or post-Kyoto Protocol). In all analyses
selection process, we obtained a total of 186 papers with information Type III sum of squares was used owing to the unequal n’s.
(Supplementary information). All the information was collected in a A correspondence analysis was undertaken to establish the re-
database where we recorded data related to the initial and final forest lationships between drivers and the different combinations of forest
surface and the years reported. Forest areas of less than 100 ha were not type and country (e.g., Atlantic forest in Argentina).

141
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

deforestation is given in Table 3. Overall, the direct deforestation dri-

vers represent almost two thirds of the causes of deforestation described
for Latin America. The expansion of agriculture was the main direct
driver of deforestation, followed by cattle grazing and infrastructure
expansion (Table 3). Geophysical variables and forestry showed inter-
mediate values, while the remaining direct drivers were not usually
identified as main causes of deforestation (Table 3). The main direct
drivers of deforestation differed considerably among the different forest
types. Thus, the most unique pattern was that of Atlantic forests, where
the agricultural and infrastructure expansions were described as the
deforestation drivers in more than 80% of cases, while the other direct
drivers had a low incidence. The other forest types showed similar
patterns, with higher incidence of agricultural expansion followed by
cattle grazing and, to a lesser extent, infrastructure expansion as the
main direct drivers. Other drivers also showed differences among forest
types: the geophysical variables had a high incidence in TDF; forestry
was more important in LW forests and TF forests than in the other forest
types; aquaculture was only determinant in MFs; and natural disasters
had the greatest impact on the deforestation of TF.
The indirect drivers represent one third of the identified causes of
deforestation. From those, the population pressure was the most im-
Fig. 1. Mean ( ± SE) deforestation rate per forest type. Different letters indicate sig- portant one, followed by markets and prices as well as by development
nificant differences according to the Tukey HSD test. AF = Atlantic forest, and agricultural policy (Table 3). Likewise, there were significant dif-
TDF = Tropical and subtropical dry forest, TF = Subtropical temperate and mixed forest, ferences among forest types. Thus, the population pressure was espe-
LW = Tropical lowland forest, TMF = Tropical montane forests.
cially described as a deforestation factor in TMF and TF, while it was
particularly rare in DTF and AF. The importance of development and
2. Results agricultural policy roughly followed this same pattern, while markets
and prices was a frequent deforestation driver in TDF and LW and
Overall, the region presents a mean annual deforestation rate of showed an especially low value in AF.
−1.14 ( ± 0.092 SE). The general analysis of deforestation rates (r) per The analysis of deforestation drivers also showed large differences
forest type revealed significant differences (F4,991 = 3.4, p = 0.009; among countries and types of forests in Latin America. The first two
Fig. 1): the deforestation rates were higher in AF, TF and TDF compared components of the correspondence analysis conducted to establish the
with LW; in contrast, TMF did not show differences with the other forest joint relationships among deforestation drivers explained approxi-
categories due to the intermediate deforestation rate observed in this mately 46% of the variance (Fig. 5). The first axis (25.7% of the var-
particular forest type. Neither the size of the area monitored in the iance explained) was characterised by the drivers of geophysical pre-
study nor the interaction of forest type and area had significant effects disposing conditions and markets and prices in the right corner and
on r (respectively, F1,991 = 2.0, p = 0.157 and F4,991 = 1.1, agricultural expansion, fire and natural disasters in the left. TDF of
p = 0.320) Despite the significance of the model, it certainly accounted Costa Rica and Venezuela, together with LW of Peru and Ecuador were
for a low predictive ability (r2 = 2%) probably because of the influence located in the right side of this axis. LW of Guatemala, Nicaragua and
of other factors not considered in this preliminary analysis (e.g., Venezuela and, to a lesser extent, Chile and Costa Rica, together with
country or period of study). Regarding the differences in deforestation TMF of Brazil and Honduras and AF of Paraguay were located in the left
rates per countries (Fig. 2, Table 1), Chile and Argentina, followed by side of this axis. A less differentiated group to the left contains mostly
Ecuador and Paraguay, were the four countries with the highest de- TMFs of Bolivia, Colombia, Guatemala, Mexico and Peru and both TDF
forestation rates on average. Costa Rica was the only country that, on and LW of Brazil and Mexico. The second axis (20.6% of variance ex-
average, showed a forest increase during the period of time covered in plained) was defined by forestry and cattle activities on one site (AF of
this study. Argentina and TF of Chile are located here, as well as TDF of Argentina,
The comparison of deforestation rates for two forest types (TDF and Costa Rica and Bolivia) and infrastructure and population pressure on
LW), with a sufficient number of literature records in different coun- the other side (with LW of Ecuador, Peru, Panama and Colombia).
tries, revealed large differences in deforestation rates among countries
during the period when the study was conducted (Table 2). In TDF 3. Discussion
forests, the larger r values were observed in Argentina and Mexico in
comparison with Costa Rica and Venezuela (Fig. 3A). Conversely, in LW The average annual deforestation rate obtained in this study is
forests, r values were significantly higher in Ecuador, intermediate in higher than all global and continental rates reported in the most recent
Brazil and México, and lower in Guatemala and Bolivia (Fig. 3B). In- reports covering this geographical area (Achard et al., 2014, 2002;
terestingly, we did not observe any significant effect of the time when FAO, 2011; Hansen et al., 2010). Since we have used a standardized
the study was conducted in TDF forests (Table 2). However, for LW, the formulae to calculate the annual rate of change of forest cover, this
significance of the interaction country x period revealed different trends effect/discrepancy is likely not explained by the different ways of re-
among countries through time (Fig. 4). Thus, Ecuador and Bolivia porting deforestation (i.e net forest cover loss) (Puyravaud, 2003).
showed, respectively, the highest and the lowest deforestation rates Because the same pattern of higher rates occurs also when analysing
with a steady pattern across the years covered by our study either countrýs deforestation rates or forest types rates, we believe the
(1953–2012), while Guatemala and especially Brazil and Mexico ex- results may differ because of the initial definition of forest types
hibited increasing rates of deforestation rates even after the Kyoto (Puyravaud, 2003). A point worth noting is that most global forest
Protocol Summit. Neither for TDF or LW the size of the area monitored studies either consider forest as a land cover class or as a forest cover
in the study or the number of years covered or their interactions had but do not discriminate different types of forests (Achard et al., 2014;
significant effects on r (Table 2). FAO, 2011; Hansen et al., 2010) while our approach, extracting and
The importance of the main direct and indirect drivers of integrating information from local studies allowed us to address forest

142
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

Fig. 2. Deforestation rates per country and forest type.

types in a more detailed approach. after Brazil, the country in Latin America with the reported highest total
From our estimates of forest loss rates, four countries stand out with deforested area between 2000 and 2012 (36, million ha), the second
higher deforestation rates, i.e., Chile, Argentina, Ecuador and Paraguay. and third countries are Argentina (4.6 million ha) and Paraguay (3.7
Indeed, this pattern can be compared to similar results reflected in re- million ha). The fact that Brazil, based on our results, reports a lower
cent regional studies using global data (Da Ponte et al., 2015), where general deforestation rate up to the year 2014 might indicate that

143
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

Table 2 higher percentage of the countrýs surface is suffering changes and

Effect of country, surface covered by the study, duration of the study, and period when shows accelerating forest loss rates (Da Ponte et al., 2015; Huang et al.,
the study was conducted (i.e., pre- or post-Kyoto Protocol) on deforestation rates of TDF
2010; Keenan et al., 2015).
and LW forests types considered separately. In bold, significant values at p < 0.05.
A second group of countries, i.e., Bolivia (2.98 million ha),
Variable TDF forests LW forests Colombia (2.5 million ha) and Mexico (2.3 million ha), have had si-
milar deforested areas during the 2000–2012 period (Da Ponte et al.,
F p F p 2015). In the case of Bolivia, our results indicate that the deforestation
Intercept 115.0 < 0.001 64.0 < 0.001 rates might be decreasing (as in Brazil). Instead, the rates are accel-
Country 8.2 0.001 2.6 0.031 erating in Colombia and Mexico. In the case of Colombia, the average
Surface 0.1 0.758 0.1 0.839 deforestation rate is higher than that reported recently for the country
Period 1.1 0.288 8.1 0.004 (Armenteras et al., 2013). Chile’s high deforestation rates based on our
Duration 0.9 0.341 0.6 0.723
results might be due to the intensification of forestry practices in this
Country × Surface 0.7 0.466 0.8 0.523
Country × Period 1.5 0.171 3.1 0.013 country (Hansen et al., 2013; Zamorano-Elgueta et al., 2015).
Country × Duration 1.0 0.305 1.1 0.207 Three Central American countries, i.e., El Salvador, Costa Rica and
Surface × Period 0.9 0.371 0.9 0.247 Panama, were reported with the lowest deforested area during
Surface × Duration 0.8 0.421 0.6 0.699
2010–2012 (Da Ponte et al., 2015). However, our results indicate a low
Period × Duration 1.1 0.355 1.4 0.180
deforestation rate or even a positive trend in forest recovery (mostly
TDF) in Costa Rica and Panama, likely associated to national forest
conservation policies (Calvo-Alvarado et al., 2009). A few reports also
show a recovery in El Salvador associated with land abandonment,
commercial expansion and urbanisation, as well as with population
displacement (Herrador Valencia et al., 2011; Keenan et al., 2015; Redo
et al., 2012).
Regarding the deforestation rates per types of forests, our results
highlighted that, in particular, two forest types, AF and TMF followed
by TDF, have been disappearing at faster rates than the more humid
LW. Indeed, AF is one of the most threatened tropical forests in the
world and, in particular, in Paraguay its condition is critical (Metzger
et al., 2009). Argentina is another key case, with large areas reported as
deforested and high deforestation rates in areas previously highlighted
as a TDF loss hotspot (Hansen et al., 2013; Keenan et al., 2015). Con-
versely, Ecuador only reports 524.600 ha deforested for that period (Da
Ponte et al., 2015). According to our results, the deforestation rate
might be accelerating; although certain reports might not be updated,
this result can be related to the deforestation in equatorial LW forests
(Mena et al., 2006; Messina and Walsh, 2001; Walsh et al., 2008). Two
other aspects can be observed for the region, in Central American
countries dry forests have recovered and deforestation rates have de-
clined, while countries like Argentina move towards a high level of
transformation i.e. the process of dry forests deforestation has shifted to
other countries. On the other hand, other types of forests different than
LW are strongly influencing deforestation dynamics in Latin America,
ie. AF and TMF. Our results consider net deforestation rates and any
potential forest regeneration that might be occurring and the rates at
which that might happen (i.e., LW or TMF typically regenerate faster
than TDF) are not considered in this study.
In relation to our second hypothesis, our results highlight the dif-
ferences between loss rates before and after the Kyoto Protocol summit
for the Latin America region, with Brazil and Mexico exhibiting a
particular pattern of increasing forest loss rates in LW forests in those
studies covering time periods mostly after 1997. This pattern suggests
that each country has its own temporal dynamic. Countries such as
Guatemala present high rates all the time, others such as Nicaragua
tend to have low rates. Moreover, countries such as Brazil, Ecuador and
Mexico, which have experienced a rapid economic growth, show in-
Fig. 3. Mean ( ± SE) deforestation rates per country. A) Tropical and subtropical dry creasing rates of deforestation. The signing of international agreements,
forests and B) Tropical lowland forests. Different letters indicate significant differences
such as the Kyoto Protocol or others, does not seem to have an im-
according to the Tukey HSD test.
mediate or direct impact on the deforestation rates. Indeed deforesta-
tion as a major source of emissions was not initially contemplated in the
deforestation rates in general havedecreased in this country as it has Kyoto Protocol. Further even when countries commit to sign this type of
also recently been emphasized (Hansen et al., 2013; Keenan et al., agreements, the implementation of them has different dynamics that
2015). However, when referring to tropical lowland forests, the de- are yet to be reflected in the reduction of rates of deforestation but
forestation rates are still high. Thus, the total deforested area per year mostly depend on the specific environmental policies at the national
in Brazil has possibly diminished by 2014 but the rates at which forest level. The market-based flexibility mechanisms (i.e., clean development
loss occurs have not. In contrast, Paraguay is a worrying case because mechanism, emissions trading etc.) embedded in the Kyoto Protocol
not only has one of the reported highest areas deforested but also a have been widely criticized, and REDD+ has been seen since 2005 as

144
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

Fig. 4. Relation between the deforestation rates per

country in Tropical lowland forests and the number
of years before or after the Kyoto Protocol Summit
(1997) when the study of reference was conducted.
Lines are drawn to illustrate the different tendencies
of each country. The length of the line reflects the
period of time available for each specific country.

Table 3 change of drivers over time (Rudel et al., 2009) and, in particular, to the
Importance (measured as the% of cases of deforestation in which each driver is involved) influence of indirect drivers of change (Meyfroidt et al., 2013), our
of the main direct and indirect drivers of deforestation, overall and in the different forest
results support the idea that direct drivers are still the most frequently
types. N, number of cases. AF = Atlantic forest, TDF = Tropical and subtropical dry
forest, TF = Subtropical temperate and mixed forest, LW = Tropical lowland forest,
reported cause of forest loss, in Latin America. This outcome may be
TMF = Tropical montane forests, MAN = Mangrove. explained by the detailed scale of our study, which is based on local
studies where a detailed scale allowed a better quantification of local
Drivers % Cases Forest types direct pressures on the forests. Thus, agricultural practices, cattle
grazing and infrastructure are the most cited causes. Regarding the
AF LW TMF TDF TF MAN
study’s third hypothesis, these causes have been reported for all the
(a) Direct drivers forest types, which highlight the extreme importance of agricultural
Geophysical variables 8.5 1.5 9.3 1.8 12.5 7.6 5.7 conversion and the use of fire in the Atlantic forest, particularly in
Infrastructure & roads 12.5 37.6 13.6 9.0 5.5 9.7 17.1
Paraguay as previously mentioned. Likewise, this pattern reflects the
Agricultural expansion 20.5 42.9 17.2 22.2 19.2 24.5 27.1
Cattle grazing 13.4 3.8 12.3 15.1 19.0 10.8 12.1 changing land use from forest to agriculture, in particular in large areas
Forestry 6.9 4.5 8.0 4.3 3.5 8.7 0.7 cultivated with soybeans, cotton and corn (Huang et al., 2010). In
Aquaculture (shrimp farms) 0.6 0.0 0.0 0.0 0.8 0.0 10.7 Brazil, the regional differences associated with different forest transi-
Fire 0.6 0.0 0.6 0.7 0.2 1.1 1.4 tion states occur and might reflect the mixing of areas being reforested
Mining 0.1 0.0 0.1 0.7 0.0 0.0 0.0
Natural disasters 0.9 0.0 0.6 0.4 0.2 4.3 0.0
with some that are under regeneration, but mostly that changes are not
constant across time and space in the country (Lira et al., 2012). In-
(b) Indirect drivers
Population pressure 13.5 8.3 14.2 21.1 5.5 17.1 15.7
direct drivers also play a role, and they vary according to the forest
Development and agricultural 7.8 0.8 8.4 14.0 7.0 7.8 1.4 types because often populations migrate to areas that have favourable
policy conditions. In the case of certain tropical dry forest areas, their bio-
Markets and prices 11.3 0.8 12.4 9.0 16.2 8.2 7.9 physical and infrastructure conditions made them more prone for tra-
Other 3.5 0.0 3.4 1.8 10.4 0.2 0.0
ditional economic activities since pre-Hispanic times (Aide et al., 2013;
Bonilla-Moheno et al., 2011; Redo et al., 2012). Similarly, TMF areas in
the most plausible way to halt deforestation and forest degradation the Andes, for instance, have been in the centre of areas with high
(Buizer et al., 2014). The pre and post REDD+ analysis of deforestation population density (Armenteras et al., 2011) whilst some of the lowest
rates will certainly be a much needed study, once more data and pub- income groups tend to migrate to LW areas towards the new colonisa-
lished reports become available. On the other hand, we are aware that tion frontiers, where natural resources are available for extraction
the way our study classified (scored) the studies based on the number of (Armenteras et al., 2011, 2006; Armenteras and Retana, 2012).
years during which the study had been conducted pre- or post- Kyoto
Protocol is a rough proxy of the impact of this policy. Yet, we believe it 4. Conclusions and recommendations
was the only way of establishing such classification and although non-
linear responses for particular studies covering both pre- and post- In order for countries to implement more effective deforestation
Kyoto years (e.g. from 1985 to 2005) where deforestation could in- policies, we believe that there is a need to reduce the uncertainty in
crease up to the Kyoto Protocol Summit, and then decrease are possible. measuring the forest loss rates per forest types, especially in countries
However, excluding the cases with the same number of pre- and post- that are still highly dynamic in terms of forest loss. Our study shows the
Kyoto years (score 0), the results did not reveal a significant influence advantage of using a standardized deforestation rate but not all studies
of this factor. report the sufficient information and many deforestation studies were
Despite it has been highlighted the need to pay more attention to the not considered because of the incomplete information. To allow future

145
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

Fig. 5. Representation of the first two axes of a correspondence analysis

carried out to establish joint relationships among deforestation drivers in the
different combinations of forest and country analysed. In red and with an
arrow, deforestation drivers (original variables): Geophy = geophysical
variables, Infras = infrastructures, Agricu = agricultural expansion,
Cattle = cattle grazing, Forest = forestry, Disast = natural disasters,
Fire = fire, Popula = population pressure, Market = markets and prices,
Develo = development and agricultural policies. In blue, the combinations of
forest type and country. Types of forests: AF = Atlantic forest,
TDF = Tropical and subtropical dry forest, TF = Subtropical temperate and
mixed forest, LW = Tropical lowland forest, TMF = Tropical montane for-
ests. Countries: Arg = Argentina, Bol = Bolivia, Bra = Brazil, Chi = Chile,
Col = Colombia, Cri = Costa Rica, Equ = Ecuador, Gua = Guatemala,
Hon = Honduras, Mex = Mexico, Nic = Nicaragua, Pan = Panama,
Par = Paraguay, Per = Peru, Ven = Venezuela. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web
version of this article.)

and better comparisons we strongly support the recommendations by Appendix A. Supplementary data
Puyravaud (2003) that all studies should report forest area, time of
measurement and rates. Focusing on the different types of forests al- Supplementary data associated with this article can be found, in the
lowed us to detect higher deforestation rates, some of which were online version, at [Link]
previously unnoticed or minimised by using a broad forest definition.
Indeed, forest decline is the result of multiple causal forces interacting, References
but certainly, this study highlights the prevalence of direct drivers that
have a strong influence at the regional level. Indeed, this study shows Achard, F., Eva, H.D., Stibig, H.-J., Mayaux, P., Gallego, J., Richards, T., Malingreau, J.-
that different types of forests are affected by different drivers, and this P., 2002. Determination of deforestation rates of the world’s humid tropical forests.
Science 80 (297), 999–1002.
knowledge is crucial in developing policies to address forest loss in each Achard, F., Beuchle, R., Mayaux, P., Stibig, H.-J., Bodart, C., Brink, A., Carboni, S.,
of them. Reducing the rates of forest loss requires the deep under- Desclée, B., Donnay, F., Eva, H.D., Lupi, A., Raši, R., Seliger, R., Simonetti, D., 2014.
standing of the local causes of change under varying conditions. Several Determination of tropical deforestation rates and related carbon losses from 1990 to
2010. Glob. Change Biol. 20, 2540–2554. [Link]
of the countries in this study do not have yet a deforestation monitoring Aide, T.M., Clark, M.L., Grau, H.R., López-Carr, D., Levy, M. a., Redo, D., Bonilla-Moheno,
system implemented, and little is known about the different dynamics M., Riner, G., Andrade-Núñez, M.J., Muñiz, M., 2013. Deforestation and reforestation
in terms of differential forest types’ losses and causes at subnational of Latin America and the caribbean (2001–2010). Biotropica 45, 262–271. [Link]
[Link]/10.1111/j.1744-7429.2012.00908.x.
levels. Hopefully, the commitments acquired by some of them in the
Armenteras, D., Retana, J., 2012. Dynamics, patterns and causes of fires in Northwestern
context described not only in the UN Framework Convention on Cli- Amazonia. PLoS One 7, e35288. [Link]
mate and their climate change policies, but also in the UN Convention Armenteras, D., Gast, F., Villareal, H., 2003. Andean forest fragmentation and the re-
presentativeness of protected natural areas in the eastern Andes. Colomb. Biol.
on Biological Diversity, the national biodiversity strategies and the
Conserv. 113, 245–256. [Link]
compliance of Aichi targets will push them to further implement ar- Armenteras, D., Rudas, G., Rodríguez, N., Sua, S., Romero, M., 2006. Patterns and causes
rangements (i.e., for the REDD+ mechanisms) such as developing a of deforestation in the Colombian Amazon. Ecol. Indic. 6, 353–368. [Link]
comprehensive monitoring approach. This could also be coupled with org/10.1016/[Link].2005.03.014.
Armenteras, D., Rodríguez, N., Retana, J., Morales, M., 2011. Understanding deforesta-
policies for strengthening community lands and different forest type tion in montane and lowland forests of the Colombian Andes. Reg. Environ. Change
governances given the different land occupation models that occur in 11, 693–705. [Link]
these countries. The recognition of the traditional and often participa- Armenteras, D., Cabrera, E., Rodríguez, N., Retana, J., 2013. National versus and regional
determinants of tropical deforestation: a case study of in Colombia. Reg. Environ.
tory land management practices by indigenous groups and local com- Change. [Link]
munities have recently gained relevance as a potential pathway to curb Bonilla-Moheno, M., Aide, T.M., Clark, M.L., 2011. The influence of socioeconomic, en-
deforestation and forest degradation in Latin America (Ceddia et al., vironmental, and demographic factors on municipality-scale land-cover change in
Mexico. Reg. Environ. Change 12, 543–557. [Link]
2015). We recommend that this approach integrate national and sub- 0268-z.
national differences in terms of both forest types coupled with local Buizer, M., Humphreys, D., de Jong, W., 2014. Climate change and deforestation: the
deforestation drivers and local governance perspectives. evolution of an intersecting policy domain. Environ. Sci. Policy 35, 1–11. [Link]
[Link]/10.1016/[Link].2013.06.001.
Calvo-Alvarado, J., McLennan, B., Sánchez-Azofeifa, A., Garvin, T., 2009. Deforestation
Acknowledgements and forest restoration in Guanacaste, Costa Rica: putting conservation policies in
context. For. Ecol. Manage. 258, 931–940. [Link]
10.035.
We thank the members of the Landscape Ecology and Ecosystem Carr, D.L., 2004. Proximate population factors and deforestation in tropical agricultural
Modelling group for their help with the search of information, parti- frontiers. Popul. Environ. 25, 585–612. [Link]
cularly Eduardo Molina, Jorge Paiba and Maria Alejandra Chadid. We 0000039066.05666.8d.
Ceddia, M.G., Gunter, U., Corriveau-Bourque, A., 2015. Land tenure and agricultural
acknowledge the financial support of the Universidad Nacional de expansion in Latin America: the role of Indigenous Peoples’ and local communities’
Colombia with its fund for strengthening international alliances and forest rights. Glob. Environ. Change 35, 316–322. [Link]
also the CYTED network IBERO-REDD+ “Network for monitoring the gloenvcha.2015.09.010.
Da Ponte, E., Fleckenstein, M., Leinenkugel, P., Parker, a., Oppelt, N., Kuenzer, C., 2015.
state of conservation and recovery of humid and dry forests in Latin Tropical forest cover dynamics for Latin America using Earth observation data: a
America in the context of avoided deforestation (CYTED P411RT0559).

146
D. Armenteras et al. Global Environmental Change 46 (2017) 139–147

review covering the continental, regional, and local scale. Int. J. Remote Sens. 36, Forest ecology and management dynamics of global forest area: results from the FAO
3196–3242. [Link] global forest resources assessment 2015 q. For. Ecol. Manage. 352, 9–20. [Link]
Eva, H.D., Belward, A.S., De Miranda, E.E., Di Bella, C.M., Gond, V., Huber, O., Jones, S., [Link]/10.1016/[Link].2015.06.014.
Sgrenzaroli, M., Fritz, S., 2004. A land cover map of South America. Glob. Change Lira, P.K., Tambosi, L.R., Ewers, R.M., Metzger, J.P., 2012. Land-use and land-cover
Biol. 10, 731–744. [Link] change in Atlantic Forest landscapes. For. Ecol. Manage. 278, 80–89. [Link]
Food and Agriculture Organization of the United Nations (FAO), 2010a. Global Forest org/10.1016/[Link].2012.05.008.
Resources Assessment 2010. Main Report. FAO Forestry Paper 163. Mather, A.S., Needle, C.L., Fairbairn, J., 1999. Environmental kuznets curves and forest
Food and Agriculture Organization of the United Nations (FAO), 2010b. Global Forest trends on JSTOR. Geography 84, 55–65.
Resource Assessment. Mena, C.F., Bilsborrow, R.E., McClain, M.E., 2006. Socioeconomic drivers of deforestation
Food and Agriculture Organization of the United Nations (FAO), 2011. State of the in the Northern Ecuadorian Amazon. Environ. Manage. 37, 802–815. [Link]
World’s Forests. (Chapter 2). org/10.1007/s00267-003-0230-z.
Geist, H.J., Lambin, E.F., 2001. What Drives Tropical Deforestation? A Meta-analysis of Messina, J.P., Walsh, S.J., 2001. Simulating land use and land cover dynamics in the
Proximate and Underlying Causes of Deforestation Based on Subnational Case Study Ecuadorian Amazon through cellular automata approaches and an integrated GIS.
Evidence (LUCC Report Series:4). LUCC International Project Office[Link] Global Environ. Change 1–13.
org/10.1098/rsbl.2008.0691. Metzger, J.P., Martensen, A.C., Dixo, M., Bernacci, L.C., Ribeiro, M.C., Teixeira, A.M.G.,
Geist, H.J., Lambin, E.F., 2002. Proximate causes and underlying driving forces of tropical Pardini, R., 2009. Time-lag in biological responses to landscape changes in a highly
deforestation. Bioscience 52, 143. [Link] dynamic Atlantic forest region. Biol. Conserv. 142, 1166–1177. [Link]
052[0143:PCAUDF][Link];2. 1016/[Link].2009.01.033.
Gibbs, H.K., Ruesch, A.S., Achard, F., Clayton, M.K., Holmgren, P., Ramankutty, N., Foley, Meyfroidt, P., Lambin, E.F., Erb, K.-H., Hertel, T.W., 2013. Globalization of land use:
J.A., 2010. Tropical forests were the primary sources of new agricultural land in the distant drivers of land change and geographic displacement of land use. Curr. Opin.
1980 and 1990. Proc. Natl. Acad. Sci. U. S. A. 107, 16732–16737. [Link] Environ. Sustain. 1–7. [Link]
10.1073/pnas.0910275107. Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V.N.,
Gomez-Peralta, D., Oberbauer, S.F., McClain, M.E., Philippi, T.E., 2008. Rainfall and Underwood, E.C., D’amico, J.A., Itoua, I., Strand, H.E., Morrison, J.C., Loucks, C.J.,
cloud-water interception in tropical montane forests in the eastern Andes of Central Allnutt, T.F., Ricketts, T.H., Kura, Y., Lamoreux, J.F., Wettengel, W.W., Hedao, P.,
Peru. For. Ecol. Manage. 255, 1315–1325. [Link] Kassem, K.R., 2001. Terrestrial ecoregions of the world: a new map of life on earth.
10.058. Bioscience 51, 933–938. [Link]
Hansen, M.C., Stehman, S.V., Potapov, P.V., Loveland, T.R., Townshend, J.R.G., DeFries, 051[0933:TEOTWA][Link];2.
R.S., Pittman, K.W., Arunarwati, B., Stolle, F., Steininger, M.K., Carroll, M., Dimiceli, Puyravaud, J.-P., 2003. Standardizing the calculation of the annual rate of deforestation.
C., 2008. Humid tropical forest clearing from 2000 to 2005 quantified by using For. Ecol. Manage. 177, 593–596. [Link]
multitemporal and multiresolution remotely sensed data. Supporting information. 00335-3.
Proc. Natl. Acad. Sci. U. S. A. 105, 9439–9444. [Link] Ramankutty, N., Gibbs, H.K., Achard, F., Defries, R., Foley, J.a., Houghton, R.a., 2007.
0804042105. Challenges to estimating carbon emissions from tropical deforestation. Glob. Chang.
Hansen, M.C., Stehman, S.V., Potapov, P.V., 2010. Quantification of global gross forest Biol. 13, 51–66. [Link]
cover loss. Proc. Natl. Acad. Sci. U. S. A. 107, 8650–8655. [Link] Redo, D.J., Grau, H.R., Aide, T.M., Clark, M.L., 2012. Asymmetric forest transition driven
1073/pnas.0912668107. by the interaction of socioeconomic development and environmental heterogeneity
Hansen, M.C., Potapov, P.V., Moore, R., Hancher, M., Turubanova, S.A., Tyukavina, A., in Central America. Proc. Natl. Acad. Sci. U. S. A. 109, 8839–8844. [Link]
Thau, D., Stehman, S.V., Goetz, S.J., Loveland, T.R., Kommareddy, A., Egorov, A., 10.1073/pnas.1201664109.
Chini, L., Justice, C.O., Townshend, J.R.G., 2013. High-resolution global maps of Rudel, T.K., Defries, R., Asner, G.P., Laurance, W.F., 2009. Changing drivers of defor-
21st-century forest cover change. Science 342, 850–853. [Link] estation and new opportunities for conservation. Conserv. Biol. 23, 1396–1405.
science.1244693. [Link]
Herrador Valencia, D., Boada i Juncà, M., Varga Linde, D., Mendizábal Riera, E., 2011. Silagy, C.A., Middleton, P., Hopewell, S., 2002. Publishing protocols of systematic re-
Tropical forest recovery and socio-economic change in El Salvador: an opportunity views. JAMA 287, 2831. [Link]
for the introduction of new approaches to biodiversity protection. Appl. Geogr. 31, Vreugdenhil, D., Meerman, J., Meyrat, A., Gómez, L.D., Graham, D.J., 2002. Map of the
259–268. [Link] Ecosystems of Central America. World Bank Washington (DC).
Holdridge, L., 1967. Life zone ecology, Rev. ed. ed. Tropical Science Center, San Jose Walsh, S.J., Messina, J.P., Mena, C.F., Malanson, G.P., Page, P.H., 2008. Complexity
Costa Rica. theory, spatial simulation models, and land use dynamics in the Northern Ecuadorian
Houghton, R.A., 2005. Aboveground forest biomass and the global carbon balance. Glob. Amazon. Geoforum 39, 867–878. [Link]
Change Biol. 11, 945–958. [Link] 011.
Houghton, R., 2012. Carbon emissions and the drivers of deforestation and forest de- Wassenaar, T., Gerber, P., Verburg, P.H.H., Rosales, M., Ibrahim, M., Steinfeld, H., 2007.
gradation in the tropics. Curr. Opin. Environ. Sustain. 4, 597–603. [Link] Projecting land use changes in the Neotropics: the geography of pasture expansion
10.1016/[Link].2012.06.006. into forest. Glob. Environ. Change 17, 86–104. [Link]
Huang, C., Goward, S.N., Masek, J.G., Thomas, N., Zhu, Z., Vogelmann, J.E., 2010. An gloenvcha.2006.03.007.
automated approach for reconstructing recent forest disturbance history using dense Zamorano-Elgueta, C., Rey Benayas, J.M., Cayuela, L., Hantson, S., Armenteras, D., 2015.
Landsat time series stacks. Remote Sens. Environ. 114, 183–198. [Link] Native forest replacement by exotic plantations in southern Chile (1985–2011) and
10.1016/[Link].2009.08.017. partial compensation by natural regeneration. For. Ecol. Manage. 345, 10–20. http://
Keenan, R.J., Reams, G.A., Achard, F., Freitas, J.V., De Grainger, A., Lindquist, E., 2015. [Link]/10.1016/[Link].2015.02.025.

147