Applied Geography 100 (2018) 101–115

Contents lists available at ScienceDirect

Applied Geography
journal homepage: www.elsevier.com/locate/apgeog

Deforestation pattern dynamics in protected areas of the Brazilian Legal T

Amazon using remote sensing data
Ana I.R. Cabrala,∗, Carlos Saitob,c, Henrique Pereirad, Anne Elisabeth Laquese
a
Forest Research Centre, School of Agriculture, University of Lisbon, Tapada da Ajuda, 1349-017, Lisbon, Portugal
b
Center for Sustainable Development, University of Brasilia, Brazil
c
Department of Ecology/Institute of Biological Sciences, University of Brasilia, Brazil
d
Federal University of Amazonas, Center for Environmental Sciences, Av. General Rodrigo Octavio Jordão Ramos, 6200 - Coroado I, Manaus, AM, 69080-900, Brazil
e
Institut de Recherche pour le Développement (IRD), 911 Avenue Agropolis, BP 64501, 34394, Montpellier Cedex 5, France

ARTICLE INFO ABSTRACT

Keywords: Forest fragmentation and deforestation are subjects of great concern in tropical regions, namely in South
Deforestation and forest degradation America and Africa, contributing to a rapid loss of tropical forest area and with serious implications for eco-
Fragmentation index system functioning and biodiversity conservation. Despite the decrease in deforestation rates in recent years, the
Remote sensing spatial data Brazilian Amazon, with the largest continuous region of tropical forest in the world, has suffered the greatest
Brazil
recorded losses, which have been contributing to continuous habitat fragmentation and a reduction in the ter-
Conservation nature
ritory occupied by Amerindian populations. In an attempt to preserve the remaining habitats and forests, Brazil
has been adopting land conservation policies, including the implementation of protected areas. Protected areas
(PAs) possess the potential to significantly reduce habitat fragmentation by conserving large, contiguous areas of
land. In order to examine how effective PAs are at conserving forest area in the Brazilian Legal Amazon, patterns
of deforestation are analyzed and compared, inside and outside the PAs, through landscape metrics calculated
using the Patch Analyst and V-LATE extensions of a Geographic Information System. Two different sources (the
Hansen Global Forest Change Dataset and the Brazilian National Institute for Space Research's (INPE) PRODES
project) of annual forest cover-loss data derived from satellite imagery at medium-to-high spatial and temporal
resolutions are compared at two-yearly intervals across 2002–2016. Additionally, fragmentation levels asso-
ciated with deforestation patterns are assessed through an index modeled using a set of uncorrelated landscape
metrics, and the associated change factors and trend are discussed. Results show that there is greater frag-
mentation in some PAs located in Mato Grosso and Pará States, especially those near the “arc of deforestation”,
and that Yanomami Indigenous Lands (YIL) are tending towards more fragmentation. Although some PAs are in
a critical condition, findings show they all actively contribute to improved conservation of the native ecosystem
and, in conjunction with sustainable management policies, will continue to help reduce or avoid forest frag-
mentation and degradation processes.

1. Introduction 2005, p. 197; Hansen et al., 2013a) report a reduction in tropical forest
on a global scale that is associated with land-use changes and has made
Tropical deforestation and forest fragmentation resulting from land- a significant contribution to the increase in CO2 emissions. According to
use changes are leading sources of concern in the research community, Pan et al. (2011), about 56% of the carbon currently held in biomass is
namely that concerned with estimating the extent of these phenomena stored in tropical forests, and land-use changes were responsible for
and their consequences for both climate change on regional and local 14–20% of global greenhouse gas emissions in 2000–2007 (Arima,
scales and the decline in global biodiversity (Arima et al., 2005, 2008; Barreto, Araújo, & Soares-Filho, 2014) and 7–14% in 2000–2005
Haddad et al., 2015; Joppa et al., 2008; Tapia-Armijos et al., 2015; (Henders et al., 2015). Deforestation influences local and regional cli-
Vedovato et al., 2016). mates, often leading to an irreversible savannization process (Malhi
Several authors (Achard et al., 2014; Baccini et al., 2012; FAO, et al., 2007) and a weakening of the affected regions, which can suffer

Corresponding author.
∗

E-mail addresses: [email protected], [email protected] (A.I.R. Cabral), [email protected] (C. Saito),

[email protected] (H. Pereira), [email protected] (A.E. Laques).

https://doi.org/10.1016/j.apgeog.2018.10.003
Received 15 May 2018; Received in revised form 6 August 2018; Accepted 15 October 2018
Available online 31 October 2018
0143-6228/ © 2018 Elsevier Ltd. All rights reserved.
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

from long periods of drought and reduced water reserves (Lovejoy & levels and patterns of forest fragmentation. Effects vary according to
Nobre, 2018). Other effects include contributions to species extinctions, edge penetration distances, spatial arrangements and time-of-persis-
loss of ecosystem services and increased habitat vulnerability, with tence of fragments (Numata & Cochrane, 2012), and imply biodiversity
long-term changes in landscape configuration that in turn lead to a loss, changes in structure and species composition, increased fire vul-
fragmentation process (Skole and Tucker, 1993; Tapia-Armijos et al., nerability and tree mortality, and canopy desiccation (Broadbent et al.,
2015). One consequence is the appearance of small, non-contiguous 2008; Cochrane & Laurance, 2002, 2008). Numata and Cochrane
fragments, in different years and with varying sizes and levels of iso- (2012) analyzed fragmentation for the BLA in seven states, using
lation, separated by a matrix of human-transformed land cover 2001–2010 PRODES data to calculate landscape metrics. Vedovato
(Broadbent et al., 2008; Haddad et al., 2015). A new ecological and et al. (2016) also assessed the status of forest fragmentation in 2014,
hydrological cycle therefore begins in each new fragment. All these using a morphological spatial pattern analysis, while Broadbent et al.
land-use transformations affect forest-dependent populations, forcing (2008) quantified rates of forest fragmentation due to deforestation and
them to change their livelihoods and traditions (Albert, De Robert, Le logging in 2005–2006. In 2010, Soares-Filho et al. used PRODES data to
Tourneau, & Laques, 2011). evaluate how efficient PAs were at reducing deforestation in
Brazil, with the largest continuous region of tropical rain forest in 1997–2008. Focusing on the spatial dimension of land-cover change,
the world, located in the Brazilian Legal Amazon (BLA), is one of the Arima, Walker, Perz, and Souza (2016) used field surveys, simulation
countries with the highest rates of forest loss (Fearnside, 2005; May and remote sensing to identify patterns of fragmentation and their re-
et al., 2016; Numata & Cochrane, 2012; Skole and Tucker, 1993; lationship with social processes.
Tyukavina et al., 2017; Vedovato et al., 2016). The long history of Improving knowledge of fragmentation processes is thus funda-
deforestation in the BLA is closely linked to colonization policies im- mental to the ability to estimate regional impacts and disturbances
plemented since the early 1960s, investments in infrastructure (i.e. (Cabral & Costa, 2017; Numata & Cochrane, 2012; Cochrane &
intensive road-building) and fiscal incentives for economic activities, Laurance, 2008), which is particularly important in PAs created with a
particularly those related with large-scale cattle-ranching and agri- conservationist goal.
culture (Carvalho et al., 2002; Kirby et al., 2006; Moran, 1993). More The aim of this study is to improve our understanding of how effective
recently, the Amazonian economy has been oriented by the demands PAs have been at protecting forest areas located in the BLA, with reference
(Nepstad et al., 2006) of the international beef and soy industries, to two-yearly intervals across 2002–2016, identifying spatial patterns and
which became important drivers of deforestation in the first half of the associated processes of change. The specific objectives are:
2000s and remain so today (Gollnow and Lakes, 2014; Laurance, 2007; To compare patterns of deforestation inside and outside PAs using
Nepstad et al., 2008; Tyukavina et al., 2017). derived remote-sensing data from two sources: Annual Forest Cover-
As part of an effort to monitor and control Amazon deforestation, Loss data produced by i) Hansen Global Forest Change Dataset (HD),
the Brazilian National Institute for Space Research (INPE) has used and ii) the Brazilian National Institute for Space Research's (INPE)
satellite data to annually map deforestation in the BLA since 1988, and PRODES project.
reports annual deforestation rates as part of the Program for To model a fragmentation index using landscape metrics, such as to
Deforestation Monitoring (PRODES) project (Almeida et al., 2016; make it possible to evaluate, for each type of PA and dataset, the level
INPE, 2017a; Maus, 2014). It also developed the Mapping of Deforested of fragmentation associated with deforestation patterns, without dis-
Areas in the Legal Amazon project (TerraClass), whose goal is to map tinguishing natural processes from anthropogenic disturbances.
land use and the spatial distribution of land cover in deforested areas To identify the main factors of change underlying landscape pat-
identified by the PRODES project (INPE, 2017b). terns.
BLA deforestation rates have varied greatly over the last fifty years, Comparing global and regional datasets allows us to evaluate the
with accelerations in the first halves of both the 1990s and 2000s, peaks advantages and disadvantages of using each type of data in terms of
in 1995 (29,059 km2/year) and 2004 (27,772 km2/year), and then a accuracy, work-time consumed and costs, and their usefulness when it
progressive fall to a historical low of 4571 km2/year in 2012 (INPE, comes to developing sustainable policies.
2017a). This decline in deforestation rates in the last decade is corro-
borated by other studies, although absolute values have varied de- 2. Data and methodology
pending on the methodology, type of forest observed and minimum
mapping unit adopted (Hansen et al., 2013b; Tyukavina et al., 2017). A 2.1. Brazilian Legal Amazon and protected areas
reversal trend occurred in 2013, with the highest value for the last eight
years recorded in 2016 (7893 km2/year), although estimates for 2017 The study area encompasses all the different types of PAs that are
(6624 km2/year) are more encouraging (INPE, 2017a). located in the BLA and help protect and conserve native ecosystems
In 2010, Brazil passed Decree (a form of legislation) no. 7,390, (Fig. 1): the various categories of Integral Protection and Sustainable-
committing to an 80% reduction in annual Amazon deforestation (to Use Conservation Units, and Indigenous Lands.
3925 km2/year) by 2020, from an average baseline of 19,625 km2/year The political-administrative region known as the BLA was created
in 1996–2005 (Gebara and Thuault, 2013; MMA, 2016). Several mea- by Decree in 1953 and comprises approximately five million km2 of
sures have been adopted in the light of this goal: stronger enforcement Brazilian territory (IBGE, 2017; Matricardi et al., 2013). It includes the
of laws, industry value-chain initiatives, expanded protected-area net- states of Acre (AC), Amapá (AP), Amazonas (AM), Pará (PA), Rondônia
working, and a robust forest-monitoring system (Nepstad et al., 2014; (RO), Roraima (RR), Mato Grosso (MT) and Tocantins (TO), as well as a
Nobre et al., 2016). Some of these strategies are actually deceptive, part of Maranhão (MA) located west of meridian 44 (Vedovato et al.,
demonstrating a decrease in illegal deforestation while legalizing new 2016). Almost the entire area (∼65%) is covered by Amazon biome
deforestation (Saito and Azevedo, 2017). Simultaneously, different characterized by primary tropical rainforest (ombrophyllous and sea-
policies based on Protected Areas (PAs) have been implemented to sonal forest), while only a small proportion (∼15%) is covered by sa-
protect tropical forests and preserve the territories of the Amerindian vanna vegetation and transitional forests, where there are grasslands
populations. and campinarana (Carreiras, Pereira, Campagnolo, & Shimabukuro,
In general, the efficiency of PAs in the fight against deforestation is 2006; FAO, 2005, p. 197; May et al., 2016). The primary tropical
consensual, but some authors (Kirby et al., 2006; Vedovato et al., 2016) rainforest is located in the central, northern and western areas, whilst
have referred to their vulnerability to both legal and illegal logging the cerrado savannas are mostly concentrated along the rims of the
activities, arguing that the influence of nearby roads, agriculture and southern and eastern areas, although there are also some isolated areas
deforestation contribute to increased degradation by fire, inducing new embedded in the tropical rainforest areas (IBGE, 2004).

102
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Fig. 1. Location of (a) the Brazilian Legal Amazon in Brazil; (b) Types of Protected Areas by level of government jurisdiction.

Vegetation has suffered severe damage as a result of political and the reserve-demarcation process was speeded up. In 2017, about 374
economic decisions. An intensive network of PAs arose in response to ILs were demarcated and identified (under the terms of Law no. 1775 of
these changes, especially after the year 2000 and the implementation of 8 January 1996), occupying about 22.8% of the BLA (ISA, 2017b).
the Brazilian National System of Nature Conservation Units (Law 9985 Insufficient information on previously defined areas and delays in the
of 18 July 2000; Decree 4340 of 22 August 2002; Decree 5746 of 5 April recognition process have contributed to overlaps between several CUs
2006) (SNUC, 2011, p. 80; Walker et al., 2009). The purpose of the PAs and ILs, and between federal and/or state CUs (Imazon, 2011). In the
is to halt the spread of deforestation by delimiting large natural and present study, we distinguished between cases in which different PA
semi-natural areas, contributing to the protection of species, habitats, categories and jurisdictions overlapped, and those with only one cate-
territories and Amerindian populations (Nogueira et al., 2018). In 2000, gory and jurisdiction. In addition, we defined a 10-km buffer zone
the area under protection represented about 10% of the BLA (Walker around each PA, based on Conama (Conselho Nacional do Meio Am-
et al., 2009). According to Soares-Filho et al. (2010), 54% of the re- biente, National Environmental Council) Resolution no. 13/1990, to
maining forests were protected by PAs by 2009. Their area increased by help us analyze whether legally non-compliant activities are taking
68% in 2004–2012, especially near the agricultural frontiers, and cur- place there. The legislation has been updated several times since 1990
rent prospects are for continued growth (Nepstad et al., 2014). Recent (Conama Resolutions nos. 428/2010, and 473/2015), but Decree
data from ISA (2017a,b) show that the actual area has increased to 99,274 of 6 June 1990, under which all activities in the 10-km-radius
about 48% of the BLA, which suggests that, if correctly managed, they areas surrounding the CUs that can affect the biota are subject to
can play an essential role in the reduction of deforestation and forest Conama rules, has remained in force. Actually, each CU can adopt its
fragmentation. Under Brazilian law, PAs include two main groups: own buffer zone, but difficulties with conducting the necessary tech-
Conservation Units (CUs), and Indigenous Lands (ILs). The National nical studies and securing approval of management plans have meant
Nature Conservation Unit System (SNUC, 2011, p. 80) classifies CUs that a 3-km radius is currently accepted (until 2020) for those CUs
into 12 categories. On the one hand, those with stricter levels of pro- without official buffer zones. Notwithstanding this varied legal frame-
tection are designed solely to preserve biological diversity and do not work, historically Brazilian policies have adopted 10 km as a reference
permit resource exploitation or human occupation; on the other, the buffer for environmental studies – a conservative value we have fol-
sustainable-use units seek to reconcile nature conservation with a sus- lowed in our work, in which the term PA is used to represent all areas
tainable use of natural resources. Conservation units can be subject to under protection as described below.
either federal, state or municipal jurisdiction. In 2017, 13.15% of the
BLA was occupied by federal CUs, with 6.81% under strict protection 2.2. Deforestation datasets
and 6.34% under sustainable-use protection (ISA, 2017a, b), while state
CUs occupied 12.45% of the region, with 2.63% under strict protection Two different sources of annual forest cover-loss data were used, at
and 9.82% under sustainable-use protection (ISA, 2017a, b). However, two-yearly intervals from 2002 to 2016: the Hansen global forest
not all CUs that are created are effectively protected, since some lack change dataset (HD), and the PRODES project (PD). The global forest
agrarian regularization and/or the drawing up and implementation of a dataset developed by Hansen et al. (2013b) entailed application of a
management plan (Nogueira et al., 2018). Indigenous Lands created decision-tree approach based on the temporal profile of spectral metrics
under federal jurisdiction seek to guarantee the land rights of tradi- derived from free Landsat imagery to produce several products focused
tional populations and the maintenance of their cultural values (see the on tree cover dynamics between 2000 and 2016 (GFC, 2018a,b; Hansen
so-called “Indian Statute”, Law no. 6001 of 19 December 1973), but on et al., b, 2013a; Linke et al., 2017; Han et al., 2017). One of these
the assumption that those populations and values only have a small products – the Forest Cover-Loss, available yearly – quantifies any tree-
impact on natural resources (Nogueira et al., 2018). In 1988, after the cover loss (percent tree cover below 25% and trees higher than 5 m)
Brazilian Constitution had guaranteed the protection of Amerindians, against a baseline of year 2000 forest cover. This means that all losses of

103
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

any type of forest (i.e. primary tropical rain forest, secondary forest and (Vector-based Landscape Analysis Tools, Lang et al. (2004)) extensions
dry tropical woodland) or deforestation dynamics (silviculture rota- and the datasets described in section 2.2. Each forest-loss map, in raster
tions, fire, selective logging, shifting cultivation and natural dis- format, was converted into vectorial format, smoothing edges to ensure
turbances) are accounted for (Hamilton and Casey, 2016; Han et al., that polygons adopted forms near the real, reduce file sizes and facil-
2017). Data is available at a spatial resolution of 1 arc-second per pixel, itate the processing step.
approximately 30 m per pixel at the equator, with no minimum map- A preliminary evaluation based on field knowledge selected an in-
ping unit defined (Landsat pixel scale) (Beyene, 2014, p. 51; GCF, itial set of four metrics, described in Table 1, as the best measures with
2018a, b; Han et al., 2017). which to characterize the fragmentation level (McGarigal et al., 2002).
The PRODES project produces annual estimates of the deforestation Metrics were calculated at the class level in order to assess the impact of

Table 1
Description of landscape composition and configuration metrics.
Structural Landscape Abbreviation Description Units/Value
category metric

Area/Density/ Number of NP Number of patches No units; > 1

Edge patches
Mean patch size MPS Average mean surface of the patches Hectares; > 0
Edge density ED Ratio between the sum of the lengths (m) of all edge m/ha; ≥0
segments and the total landscape area (m2), multiplied = 0; No edge, consisting of a single patch
by 10,000 (convert to hectares)
Isolation/ Mean proximity PROX Sum of the ratio between area and inter-patch distance No units; ≥0
Proximity index for all patches within a predefined buffer distance around = 0; patch has no neighbors within the specific search radius;
a patch (a radius of 1000 m was chosen as the buffer) > 0; increases as the neighborhood (defined by the specified
search radius) is increasingly occupied by patches of the same
type and as those patches become closer and more contiguous
(or less fragmented) in distribution

rate and the spatial extent of deforestation since 1988 at the BLA level. human occupation and land-use dynamics.
The methodology1 applies a linear spectral mixing model to Landsat Several studies have focused on the efficiency of using a set of
data to obtain the shade, soil and vegetation endmember proportions metrics in the landscape fragmentation analysis to avoid data re-
used to classify the five classes of PRODES classification (Morton & dundancy, rather than using all available metrics (Ghosh et al., 2012;
DeFries, 2005): clouds, non-forest, previous deforestation, new defor- Tian et al., 2011). To ensure that a set of uncorrelated metrics was
estation, and forest classes. It only detects large-scale deforestation of chosen in this case, we used the Pearson correlation coefficient criteria
disturbed and undisturbed primary forest – old-growth forests of dense and a multivariate statistic method based on Principal Component
humid tropical forest biome – adopts a minimum mapping unit of Analysis (PCA).
6.25 ha, and provides data with a spatial resolution of 60 m.
For each dataset, data were extracted for all types of PA and a 10-km
2.4. Modelling a fragmentation index
surrounding buffer. In addition, a minimum mapping unit of 1 ha was
adopted in accordance with the Brazilian forest definition based on the
Fragmentation levels were evaluated through an index modeled
United Nations Framework Convention on Climate Change (UNFCCC)
using the landscape metrics and the Principal Component Analysis
(Sasaki and Putz, 2009): an area of land greater than 1 ha, with more
(PCA) method. First, the four metrics were standardized based on the
than 30% canopy cover and a minimum tree height of 5 m. The
standard deviation model to make it possible to compare values of
UNFCCC defines forest as an area of land 0.05–1 ha in size, of which
different variables. For each dataset, all standardized metric values
more than 10–30% is covered by tree canopy, but countries partici-
calculated inside PAs, for all years, were used to model the fragmen-
pating in the program are free to choose their own forest definition
tation index equation. The purpose was to obtain a single equation per
within those ranges (Sasaki and Putz, 2009). As such, sixteen forest-loss
dataset that could represent all fragmentation variations in the period
maps were obtained for each dataset, eight inside and eight outside
under analysis. The same procedure was followed for buffer areas.
PAs. Each PA and the corresponding buffer was labelled with its own
Based on Tian et al. (2011) and on Cumming and Vernier (2002), the
attribute to enable analysis. The term “annual forest loss” appears,
PCA method was executed, using IBM SPSS software (IBM Corporation
along the text, since it represents the values of forest cover loss in each
Released, 2017), with a varimax rotation criterion, to obtain the ei-
year analyzed (2002-2004-2006-2008-2010-2012-2014-2016). In fact
genvalues (which explain the variance of the metrics), which were used
we use annual forest cover-loss data at two-yearly intervals across
as the weight attributable to each component. The varimax rotation
2002–2016.
criterion was used to aid in the interpretation of component loadings
(the coefficients of individual variables in the principal components)
2.3. Metric selection for spatial deforestation pattern analysis (Cumming & Vernier, 2002) since it attempts to prevent the variables of
having high factor loadings for various components (Rossoni et al.,
Analyzing spatial patterns of deforestation in order to detect and 2016).
quantify temporal changes requires the collection of information on the The variables (metrics) with higher loadings in each component
composition and configuration of patches (Molina et al., 2015). Our (Table 3) were retained and used to build the fragmentation index. The
deforestation analysis was based on landscape metrics calculated in the first two components, with eigenvalues greater than one, always ex-
Geographic Information System ArcGIS 10.5.0.6491 (Environmental plain the variance above 78.1%, and they represent the characteristics
Systems Research Institute (ESRI), 2016) using the Patch Analyst ver- of the four metrics (Table 2). According to Kaiser (1960), principal
sion 5.2.0.16 (Rempel et al., 2012) and V-LATE version 2.0 beta components are sufficiently reliable if eigenvalues are greater than one.
Principal component (PC) 1 always represented the relationships be-
tween patches of deforestation (NP and ED), while PC 2 was related
1
http://www.obt.inpe.br/prodesdigital/metodologia.html. with the degree of proximity between patches and their size (MPS and

104
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Table 2
Factor loadings before and after the varimax rotation criterion in the PCA method.
Dataset Factor Initial eigenvalues Extraction sums of loadings squared Rotation sums of loading squared

Total %Variance %Cumulative Total %Variance %Cumulative Total %Variance %Cumulative

Hansen inside PAs 1 2.230 55.746 55.746 2.230 55.746 55.746 2.045 51.120 51.120
2 1.066 26.652 82.399 1.066 26.652 82.399 1.251 31.278 82.399
3 0.670 16.750 99.149
4 0.034 0.851 100.000
Hansen outside PAs 1 2.039 50.972 50.972 2.039 50.972 50.972 1.965 49.124 49.124
2 1.085 27.130 78.101 1.085 27.130 78.101 1.159 28.977 78.101
3 0.848 21.192 99.293
4 0.028 0.707 100.000
PRODES inside PAs 1 2.008 50.190 50.190 2.008 50.190 50.190 1.926 48.142 48.142
2 1.394 34.838 85.028 1.394 34.838 85.028 1.475 36.887 85.028
3 0.529 13.231 98.260
4 0.070 1.740 100.000
PRODES outside PAs 1 1.979 49.470 49.470 1.979 49.470 49.470 1.951 48.776 48.776
2 1.239 30.964 80.434 1.239 30.964 80.434 1.266 31.658 80.434
3 0.740 18.492 98.925
4 0.043 1.075 100.00

In bold are highlighted the eigenvalues higher than 1.

PROX) (Table 3). patch area, and PROXi is the standardized value of the proximity index.
The fragmentation index, here designated Spatial Deforestation Considering the different Amazonian landscapes (Dubreuil, Laques,
Fragmentation Index (SDFI), was constructed, based on the eigenvalues Nédélec, Arvor, & Gurgel, 2008), we visually interpreted all metrics in
of the two principal components (Table 2) and the corresponding order to understand how they could be spatially combined to obtain
loading values of each metric (Table 3), as follows: each degree of fragmentation (Fig. 2).
The Jenks natural breaks method was used to divide median index
SDFIi = E i1 ×(S1×NPi + S2 ×EDi )+ E i2 ×(S3×MPSi + S4 ×PROXi) values, inside and outside PAs, into four levels of fragmentation: Low,
Median, High, and Very High. Additionally, we analyzed the SDFI
where i is the PA, Ei1 and Ei2 are the eigenvalues of the first and second trend, based on the slope parameter of the linear regression equation,
components respectively, S is the loading value of each metric, NPi is which was divided into three classes also using Jenks natural breaks:
the standardized value of the number of patches, EDi is the standardized Decrease, Stable, and Increase.
value of the edge density, MPSi is the standardized value of the mean

Table 3
Metrics loading on each factor (component) in the PCA method.
Dataset Metrics Component

1 2

Hansen inside PAs NP 0.974 0.069

MPS −0.053 0.896
ED 0.972 0.161
PROX 0.387 0.647
Hansen outside PAs NP 0.992 0.035
MPS 0.003 0.789
ED 0.983 0.131
PROX 0.123 0.719
PRODES inside PAs NP 0.982 0.019
MPS 0.084 0.853
ED 0.977 0.103
PROX 0.022 0.859
PRODES outside PAs NP 0.989 −0.02
MPS 0.028 0.796
ED 0.985 0.090
PROX 0.041 0.790

105
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Fig. 2. Spatial relationship between the four landscape metrics and the degree of fragmentation.

3. Results and discussion outside = 2.11%), as was the mean annual rate (HD: inside = 0.10%,
outside = 0.36%, PD: inside = 0.05%, outside = 0.26%). These values
3.1. Deforestation dynamic analysis could be due to different forest definitions and methodologies adopted
by each dataset, as mentioned in section 2.2.
Time series (2002–2016) of Hansen (H) and PRODES (P) datasets The higher values estimated by HD may be explained by the fact
(D) were used to obtain the annual and cumulative forest loss inside and that it counts losses linked to all forest types and dynamics, whereas PD
outside PAs (Fig. 3), as well as the annual and cumulative extent of only considers primary forest. Annual forest loss increased in
forest loss (Table 4). 2002–2004 (except HD outside), then became relatively stable until
The trends in the extent of annual forest loss were different for HD 2014, before rising again in 2014–2016. This was the case for both
and PD: the former increased from 130,021.9 ha to 707,263.0 ha, and datasets, but values inside PAs were lower than outside. Azevedo and
from 479,742.9 ha to 702,947.95 ha, inside and outside PAs respec- Saito (2013) analyzed the total extent of the deforestation that occurred
tively, while the latter decreased from 250,057.1 ha to 97,053.3 ha, and in Mato Grosso State and was authorized by the Rural Property En-
465,762.3 ha to 201,390.6 ha. However, both datasets presented higher vironmental Licensing System (SLAPR) in 2000–2007. They found that
forest loss outside PAs. an increase in authorized deforestation with a high impact on total
The cumulative forest loss over the eight years was higher in HD deforestation area was observed precisely in 2002–2004. 2004 saw the
(inside = 0.81%, outside = 2.93%) than in PD (inside = 0.43%, implementation of the Action Plan for the Prevention and Control of

106
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Fig. 3. Annual and cumulative forest loss inside and outside Protected Areas in the Brazilian Legal Amazon for Hansen Global Dataset (HD) and PRODES Dataset
(PD).

Deforestation in the Legal Amazon (PPCDAm), which contributed to Higher cumulative forest loss was observed, inside PAs, in Pará and
slow down the deforestation until 2012 (MMA, 2011). The largest re- Mato Grosso States, and outside PAs, in Acre, Rondônia and Amazonas
ductions were observed in Mato Grosso State, which was the greatest States, in the priority region referred to above (Table 5 and Fig. 5).
contributor to annual deforestation in the BLA (Azevedo & Pasquis, Araújo et al. (2017) highlighted some of these areas as being among
2007). Since 2003, the federal government has prioritized the creation the fifty PAs with higher levels of deforestation in the BLA. Although
of PAs to the south of the River Amazon, as a clear strategy for creating less risk of degradation was observed in Maranhão and Tocantins
a barrier to the advance of the deforestation boundary in as-yet pre- States, they presented high levels of forest loss in HD, which may
served regions. Two major regions have been considered priority areas: suggest higher levels of environmental degradation in the future.
the Porto Velho (RO) – Lábrea (AM) – Rio Branco (AC) triangle, and the According to Araújo et al. (2017, p. 92), about 87% of the defor-
region known as “Terra do Meio” (literally “Middle Earth”) in south- estation in the Legal Amazon in 2012–2015 occurred in Pará and
central Pará, at the intersection of the Rivers Iriri and Xingu. These two Rondônia States, usually near large infrastructures like road networks
regions together include three ILs and two state and four federal UCs and hydroelectric installations. ISA (2015) referred the impact the
(for a total of 8.4 million ha) and are currently subject to strong social latter factors have had on CUs and ILs and on the definition of new PA
pressure on forests and increases in deforestation around PAs. boundaries, in the sense that they both promote occupation. In the
The level of protection (Integral Protection (IP), Sustainable-Use present study, the SP (Environmental Protection Area (EPA), National
Protection (SP), Environmental Protection Area (EPA), or Indigenous Forest (NF), Extractive Reserve (ER)) and IL categories presented higher
Lands (IL)) seems to influence the degree of forest loss (Fig. 4). levels of forest loss, while in the IP category only one PA (National Park
Given their high level of degradation and deforestation as noted by – NP) did so. Soares-Filho et al. (2010) said that three categories of PA
Araújo, Barreto, Baima, and Gomes (2017), the EPAs were analyzed (Indigenous Lands, Integral Protection, and Sustainable-Use) appeared
separately from the SP units. An increase in cumulative forest loss in- to have an inhibiting effect on deforestation. In this respect, they
side and outside EPAs strongly indicated that this category is incapable especially noted the Xingu, Jarina, Menkragnotí, Baú and Kayapó ILs in
of protecting forest. On the contrary, ILs located in regions with high central Mato Grosso and Pará States. However, our study indicated the
levels of human pressure and a greater increase in forest loss in sur- existence of substantial loss of forest in three of these areas – Xingu, Baú
rounding areas continued to experience low proportions of forest loss and Kayapó. The disturbances in Xingu and Baú are mainly due to
within their perimeters. anthropic actions, but in Kayapó they are the result of natural

Table 4
Extent of annual and cumulative forest loss inside and outside Protected Areas in the Brazilian Legal Amazon in 2002–2016.
Dataset Year Annual Annual Cumulative Cumulative
forest loss forest loss forest loss (ha) forest loss (%)
(ha) (%)

Hansen inside PAs 2002 130 021.9 0.06 130 021.9 0.06
2016 707 263.0 0.31 1 868 012.4 0.81
Hansen outside PAs 2002 479 742.9 0.46 479 742.9 0.47
2016 702 948.0 0.68 3 017 426.8 2.93
PRODES inside PAs 2002 250 057.1 0.11 250 057.1 0.11
2016 97 053.3 0.04 1 004 493.3 0.43
PRODES outside PAs 2002 465 762.3 0.45 465 762.3 0.45
2016 201 390.6 0.20 2 174 398.4 2.11

107
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Fig. 4. Annual and cumulative forest loss considering different types of protection area: Conservation Units (CUs) (three groups: IP-Integral Protection, SP-
Sustainable-Use Protection, and EPA-Environmental Protection Area), and IL-Indigenous Lands.

Table 5
Protected Areas with higher forest loss.
Dataset State Type of PA Jurisdiction Protected Area

Hansen inside PAs Pará Environmental Protection Area State Triunfo do Xingu
National Forest Federal Jamanxim
Extractive Reserves Federal Tapajós Arapiuns
Mato Grosso Indigenous Land Federal Xingu Park
Tocantins Environmental Protection Area State Ilha do Bananal/Cantão
Maranhão Environmental Protection Area State Baixada Maranhense
Indigenous Land Federal Arariboia
Hansen outside PAs Pará Environmental Protection Area State Triunfo do Xingu
Extractive Reserves Federal Tapajós Arapiuns
Indigenous Land Federal Kayapó
Baú
Trincheira Bacajá
Apyterewa
Mato Grosso Indigenous Land Federal Xingu Park
Maranhão Environmental Protection Area State Baixada Maranhense
Indigenous Land Federal Araribóia
Rondônia Indigenous Land Federal Eru-Eu-Wau-Wau
Amazonas National Forest Federal Iquiri
National Park Federal Mapinguari
Acre Extractive Reserves Federal Chico Mendes
PRODES inside PAs Pará Environmental Protection Area State Triunfo do Xingu
National Forest Federal Jamanxim
Indigenous Land Federal Apyterewa
(continued on next page)

108
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Table 5 (continued)

Dataset State Type of PA Jurisdiction Protected Area

PRODES outside Pará National Forest Federal Jamanxim

PAs Indigenous Land Federal Baú
Trincheira Bacajá
Parakanã
Mato Grosso Indigenous Land Federal Xingu Park
Rondônia Indigenous Land Federal Eru-Eu-Wau-Wau
Acre Extractive Reserves Federal Chico Mendes
Amazonas National Forest Federal Iquiri
National Park Federal Mapinguari

Fig. 5. Cumulative deforestation for 2002–2016. Interval values, in hectares, defined by Jenks natural breaks: Low (0, 7857), Median (7857, 28 724), High (28 724,
81 936) and Very High (81 936, 258 510).

processes, in that the region is characterized by numerous exposed identify some hotspots that need additional conservative measures;
rocky formations. however, by presenting higher values, it can also lead to additional and
possibly unnecessary limits on the implementation of policies oriented
3.2. Spatial deforestation patterns and configuration towards economic development.
The smaller number of fragments (deforested patches) shows that
The HD presented greater fragmentation than the PD, with a higher the fragmentation process is smoother within PAs, but in both cases
number of fragments, lower mean patch size, and greater edge density most patches were between 100 and 10,000 ha in size (Figs. 6 and 7).
and proximity, thereby increasing the reported fragmentation process
(Fig. 6), as expected given their higher forest loss. In a conservative
approach, HD can be useful as a preventive tool, since it can help

109
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

3.3. Fragmentation index analysis

The index we used was designed to evaluate forest fragmentation in

PAs and their surroundings and identify possible critical areas where PA
regulations are insufficient to protect forest on their own and additional
conservation measures may be needed. Fig. 8 shows the average index
values for 2002–2016. Most of the areas presented a low or median risk
of fragmentation. Only two PAs located in Pará State (EPA-Triunfo do
Xingu and NF-Jamanxim for both HD and PD, and IL-Kayabi for PD
only), and one each in Mato Grosso (IL-Xingu Park for HD), Maranhão
(EPA-Baixada Maranhense for HD), Tocantins (EPA-Ilha do Bananal/
Cantão for HD), and Acre (ER-Chico Mendes for PD) States presented
higher levels of risk. Outside PAs, two areas were at risk in Pará State
(EPA-Triunfo do Xingu and IL-Parakanã for HD, and IL-Kayapó and
Parakanã for PD), along with three in the Porto Velho (RO) – Lábrea
(AM) – Rio Branco (AC) (ER-Chico Mendes and NP-Mapinguari, and IL-
Eru-Eu-Wau-Wau, respectively for both HD and PD) triangle and just
one in Maranhão State (EPA-Baixada Maranhense for HD). According to
Fig. 6. Average class metrics for each Hansen Global dataset (HD) and PRODES Mascarenhas et al. (2017), deforestation in ER-Chico Mendes is asso-
dataset (PD) inside and outside Protected Areas. ciated with cattle-ranching, and with forest fires due to extreme
drought events. These fires occur not only in regions with greater
human pressure, but also in isolated areas without terrestrial access,
reflecting the forest's great vulnerability to extreme drought.
Most of the at-risk areas are in the arc of deforestation, which en-
compasses southeastern Maranhão, northern Tocantins, southern Pará,
northern Mato Grosso, Rondônia, southeastern Amazonas, and the
southeastern areas of Acre. This is a region where cattle-ranching,

Fig. 7. Number of deforested fragments by size class for the base year 2002 and cumulative years.

110
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Fig. 8. Average fragmentation index for 2002–2016. Interval values (no units) defined by Jenks natural breaks: Low (−2.37, 1.52), Median (1.52, 10.07), High
(10.07, 33.20) and Very High (33.20, 124.39).

small-scale farming and, in the last fifty years, large-scale soy farming Nevertheless, we can see that the apparent stability of the frag-
have been playing important roles in the clearing of Amazonian forests mentation process is geographically unequal, suggesting that there are
(Arvor, Meirelles, Dubreuil, Bégué, & Shimabukuro, 2012; Colson, some hotspots in the Northern region (mainly in the Yanomami
Bogaert, & Ceulemans, 2011). The increase in these activities has con- Indigenous Land (YIL) in Roraima State) and the arc of deforestation
tributed to a denser road network, which many authors have suggested where fragmentation is tending to increase. Although the arc is already
is one of the two main factors behind the increase in deforestation, the a well-known deforestation hotspot, the rising rates of forest fragmen-
other being population density (Colson et al., 2011). tation in the YIL confirm an increase in illegal mining activity (Nilsson
As we mentioned earlier, our deforestation analysis showed an in- & Fearnside, 2011), which has been continuously resisted by the local
creased trend towards fragmentation in EPAs and SPs (Fig. 9). indigenous community and government. These mining activities take
EPAs (included in the SP group) allow people to live inside PAs, but place both in water and on land and require the support of light-aircraft
are considered the weakest category in Brazil's environmental policy landing strips in the forest, which themselves cause additional defor-
when it comes to the enforcement of nature protection. They are usually estation and a loss of forest humidity. Nilsson and Fearnside (2011) call
characterized by larger areas, including urban settlements and even goldmining activities the main contributor to forest clearing, especially
large cities, whereas the SP group, excluding EPAs, is designed to foster following the invasion of the YIL by goldminers in 1994–1995, and
the sustainability of traditional communities and their ancestral prac- highlight their impact on Yanomami communities.
tices. In general, fragmentation seems to have been stable in the period
under analysis (Fig. 10).

111
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Fig. 9. Annual and cumulative fragmentation index values. Types of protection area: Conservation Units (CUs) (three groups: IP-Integral Protection, SP-Sustainable-
Use Protection, and EPA-Environmental Protection Area), IL-Indigenous Lands.

112
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Fig. 10. Fragmentation trends, 2002–2016. Interval values (no units) defined by Jenks natural breaks: Decrease (−2.76, −0.37), Stable (−0.37, 1.03), Increase
(1.03, 5.71).

4. Conclusions highlighted – a result that contrasts with the idea given by deforestation
data or fragmentation indices. Fragmentation trends identified this as
Given the increased forest loss in the Brazilian Legal Amazon in the an area that needs to be carefully considered in more detail, with a new
last few decades and the significant impact on the life of people who are look at the socio-environmental conflicts and hazardous drivers that
dependent on the forest, it is important to obtain information on the may be heralding a disaster in the near future. Although global data
past and present states of forests that will help policymakers implement cannot replace regional information (with its greater accuracy), it does
rules which permit sustainable management. Such data make it possible help build a reliable historical profile of the state of forests when no
to estimate the effectiveness of the Protected Areas (PAs) that are cre- regional data exists.
ated with the goal of preserving forest cover. Evaluating changes on a In addition, our analysis shows that PAs play an important role in
regional scale in the BLA is difficult, expensive and time-consuming, the sustainability of the forest ecosystem in the BLA and are a powerful
because it implies the analysis of large amounts of data. It is thus im- weapon in the fight to reduce deforestation and forest degradation – a
portant to find new types of data that provide accurate and up-to-date conclusion that matches those of a number of authors (Bebber & Butt,
information while minimizing costs. Within this context, a comparative 2017; Lee et al., 2007; Nicolle and Leroy, 2017; Soares-Filho et al.,
analysis of the most commonly used datasets is useful to an under- 2010). However, care is needed to ensure that local people are not
standing of both limits and possibilities. In the present study, a global adversely affected by the implementation of conservative measures
dataset produced by automatic methods was compared with a regional (Andersen et al., 2017). The real challenge is to achieve two simulta-
dataset developed by the country in question using a specific time- neous compromises: between conservation measures and anti-poverty
consuming methodology. Instead of resorting to a traditional analysis programs that enable families to increase their incomes; and between a
based on forest patterns, we used forest-loss patterns. Regardless of the reduction in poverty and the preservation of the forest and respect for
data source, deforestation and fragmentation analysis revealed a gen- the right of Amerindians to use their territory.
erally stable trend throughout the BLA except for some hotspots in the
arc of deforestation, where the trend is a rising one. A mixed analysis Acknowledgments
combining deforestation, fragmentation index and trends provides
some interesting insights, in that the Yanomami Indigenous Land was The projects leading to this work have received funding from the

113
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

European Union's Horizon 2020 Research and innovation programme Colson, F., Bogaert, J., & Ceulemans, R. (2011). Fragmentation in the Legal Amazon,
under the Marie Skłodowska - Curie grant agreement No. 691053 Brazil: Can landscape metrics indicate agricultural policy differences? Ecological
Indicators, 11(5), 1467–1471. https://doi.org/10.1016/j.ecolind.2010.12.020.
(H2020-MSCA-RISE-2015 ODYSSEA project) and from the International Cumming, S., & Vernier, P. (2002). Statistical models of landscape pattern metrics, with
Cooperation Program GuyAmazon-Edital n.022/2014 (IRD/UFAM), applications to regional scale dynamic forest simulations. Landscape Ecology, 17(5),
which funded the SINBIOSE “Système d'INdicateurs de BIOdiversité à 433–444. https://doi.org/10.1023/A:1021261815066.
Dubreuil, V., Laques, A.-E., Nédélec, V., Arvor, D., & Gurgel, H. (2008). Paysages et fronts
l'uSage des actEurs: Biodiversité terrestre et aquatique (Amazone & pionniers amazoniens sous le regard des satellites: l'exemple du Mato Grosso. L'Espace
Oyapock)” project. CEF which is a research unit funded by Fundação Géographique, 37(1), 57–74. https://doi.org/10.3917/eg.371.0057.
para a Ciência e a Tecnologia I.P. (FCT), Portugal (grants UID/AGR/ Environmental Systems Research Institute (ESRI (2016). ArcGIS desktop 10.5, version
10.5.0.6491. Redlands, CA.
00239/2013). FAO (2005). Computerized data gathering and networking as a control and monitoring system
for the improvement of and reporting on forest management in the Amazon: The case of
References Brazil. Forest management working paper FM/27. Rome, Italy: FAO. http://www.fao.
org/tempref/docrep/fao/009/j5416e/j5416e00.pdf, Accessed date: 6 August 2018.
Fearnside, P. M. (2005). Deforestation in brazilian amazonia: History, rates and con-
Achard, F., Beuchle, R., Mayaux, P., Stibig, H.-J., Bodart, C., Brink, A., et al. (2014). sequences. Conservation Biology, 19(3), 680–688. https://doi.org/10.1111/j.1523-
Determination of tropical deforestation rates and related carbon losses from 1990 to 1739.2005.00697.x.
2010. Global Change Biology, 20(8), 2540–2554. https://doi.org/10.1111/gcb.12605. Gebara, M. F., & Thuault, A. (2013). GHG mitigation in Brazil's land use sector: An in-
Albert, B., De Robert, P., Le Tourneau, F.-M., & Laques, A.-E. (2011). From amerindian troduction to the current national policy landscape. Working Paper. Washington, DC:
territorialities to “indigenous lands” in the brazilian Amazon: The Yanomami and World Resources Institute. https://www.wri.org/sites/default/files/ghg-mitigation-
Kayapó cases. Catherine aubertin and estienne rodary. Protected areas, sustainable brazil-land-use-sector.pdf, Accessed date: 6 August 2018.
land? Ashgate, 123–141. GFC (2018a). Global forest change 2000-2014. https://earthenginepartners.appspot.com/
Almeida, C. A., Coutinho, A. C., Esquerdo, J. C. D. M., Adami, M., Venturieri, A., Diniz, C. science-2013-global-forest/download_v1.2.html, Accessed date: 6 August 2018.
G., et al. (2016). High spatial resolution land use and land cover mapping of the GFC (2018b). Global forest change 2000-2014. https://earthenginepartners.appspot.com/
Brazilian Legal Amazon in 2008 using Landsat-5/TM and MODIS data. Acta science-2013-global-forest/download_v1.4.html, Accessed date: 6 August 2018.
Amazonica, 46(3), 291–302. https://doi.org/10.1590/1809-4392201505504. Ghosh, A., Munshi, M., Areendran, G., & Joshi, P. K. (2012). Pattern space analysis of
Andersen, L. E., Groom, B., Killick, E., Ledezma, J. C., Palmer, C., & Weinhold, D. (2017). landscape metrics for detecting changes in forests of Himalayan Foothills. Asian
Modelling land use, deforestation and policy: Optimization-heterogeneous agent Journal of Geoinformatics, 12(1) 12 pp.
model with application to the bolivian Amazon. Ecological Economics, 135, 76–90. Gollnow, F., & Lakes, T. (2014). Policy change, land use, and agriculture: The case of soy
https://doi.org/10.1016/j.ecolecon.2016.12.033. production and cattle ranching in Brazil, 2001-2012. Applied Geography, 55, 203–211.
Araújo, E., Barreto, P., Baima, S., & Gomes, M. (2017). Unidades de conservação mais https://doi.org/10.1016/j.apgeog.2014.09.003.
desmatadas da Amazônia Legal (2012-2015). Belém: Instituto do Homem e Meio Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, A., Holt, R. D., et al.
Ambiente da Amazônia978-85-86212-89-5. (2015). Habitat fragmentation and its lasting impact on Earth's ecosystems. Science
Arima, E. Y., Barreto, P., Araújo, E., & Soares-Filho, B. (2014). Public policies can reduce Advances, 1(2), 1. https://doi.org/10.1126/sciadv.1500052 e1500052.
tropical deforestation: Lessons and challenges from Brazil. Land Use Policy, 41, Hamilton, S. E., & Casey, D. (2016). Creation of high spatio-temporal resolution global
465–473. https://doi.org/10.1016/j.landusepol.2014.06.026. database of continuous mangrove forest cover for 21st century (CGMFC-21). Global
Arima, E. Y., Walker, R. T., Perz, S. G., & Caldas, M. (2005). Loggers and forest frag- Ecology and Biogeography, 25(6), 729–738. https://doi.org/10.1111/geb.12449.
mentation: Behavioral models of road building in the Amazon basin. Annals of the Han, X., Josse, C., Young, B. E., Smyth, R. L., Hamilton, H. H., & Bowles-Newark, N.
Association of America Geographers, 95(3), 525–541. https://doi.org/10.1111/j.1467- (2017). Monitoring national conservation progress with indicators derived from
8306.2005.00473.x. global and national datasets. Biological Conservation, 213(B), 325–334. https://doi.
Arima, E. Y., Walker, R. T., Perz, S., & Souza, C., Jr. (2016). Explaining the fragmentation org/10.1016/j.biocon.2016.08.023.
in the brazilian amazonian forest. Journal of Land Use Science, 11(3), 257–277. Hansen, M. C., Potapov, P. V., Moore, R., Hancher, M., Turubanova, S. A., Tyukavina, A.,
https://doi.org/10.1080/1747423X.2015.1027797. et al. (2013a). High-resolution global maps of 21st-century forest cover change.
Arima, E. Y., Walker, R., Sales, M., Sousa, C., Jr., & Perz, S. G. (2008). The fragmentation Science, 342(6160), 850–853. https://doi.org/10.1126/science.1244693.
of the space in the Amazon Basin: Emergent road networks. Photogrammetric Hansen, M. C., Potapov, P. V., Moore, R., Hancher, M., Turubanova, S. A., Tyukavina, A.,
Engineering and Remote Sensing, 74(6), 699–709. et al. (2013b). High-resolution global maps of 21st-century forest cover change.
Arvor, D., Meirelles, M., Dubreuil, V., Bégué, A., & Shimabukuro, Y. E. (2012). Analyzing Science, 342(6160), 850–853. supplementary materials http://science.sciencemag.
the agricultural transition in Mato Grosso, Brazil, using satellite-derived indices. org/content/suppl/2013/11/14/342.6160.850.DC1ttp://science.sciencemag.org/
Applied Geography, 32(2), 702–713. https://doi.org/10.1016/j.apgeog.2011.08.007. content/suppl/2013/11/14/342.6160.850.DC1, Accessed date: 6 August 2018.
Azevedo, A. A., & Pasquis, R. (2007). Da abundância do agronegócio à Caixa de Pandora Henders, S., Persson, U. M., & Kastner, T. (2015). Trading forests: Land use change and
ambiental: A retórica do desenvolvimento (in) sustentável do Mato Grosso (brasil). carbon emissions embodied in production and exports of forest-risk commodities.
Revista Internacional de Desenvolvimento Local, 8(2), 183–191. Environmental Research Letters, 10(12), https://doi.org/10.1088/1748-9326/10/12/
Azevedo, A. A., & Saito, C. H. (2013). O perfil dos desmatamentos em Mato Grosso, após 125012 13 pp.
implementação do licenciamento ambiental em propriedades rurais. Cerne, 19(1), IBGE (2004). Mapa de biomas e vegetação. https://ww2.ibge.gov.br/home/presidencia/
111–122. https://doi.org/10.1590/S0104-77602013000100014. noticias/21052004biomashtml.shtm, Accessed date: 6 August 2018.
Baccini, A., Goetz, S. J., Walker, W. S., Laporte, N. T., Sun, M., Sulla-Menashe, D., et al. IBGE (2017). Amazônia legal. https://www.ibge.gov.br/geociencias-novoportal/
(2012). Estimated carbon dioxide emissions from tropical deforestation improved by informacoes-ambientais/geologia/15819-amazonia-legal.html, Accessed date: 6
carbon-density maps. Nature Climate Change, 2, 182–185. August 2018.
Bebber, D. P., & Butt, N. (2017). Tropical protected areas reduced deforestation carbon IBM Corporation Released (2017). IBM SPSS statistics for windows, version 25.0. Armonk,
emissions by one third from 2000-2012. Scientific Reports, 7(14005), https://doi.org/ NY: IBM Corp.
10.1038/s41598-017-14467-w 7 pp. Imazon (2011). Protected areas in the brazilian Amazon: Challenges and opportunities.
Beyene, D. L. (2014). Assessing the applicability of global forest cover change datasets to Imazon/Instituto Socioambiental (ISA) CDD-304.2709811.
national REDD+ forest monitoring: The case of Ethiopia. Centre for geo-information, INPE (2017a). Projeto PRODES: Monitoramento da Floresta Amazônica Brasileira por satélite.
minor thesis report GIRS-2014-29. Wageningen University. Instituto Nacional de Pesquisas Espaciaishttp://www.obt.inpe.br/OBT/assuntos/
Broadbent, E. N., Asner, G. P., Keller, M., Knapp, D. E., Oliveira, P. J. C., & Silva, J. N. programas/amazonia/prodes, Accessed date: 6 August 2018.
(2008). Forest fragmentation and edge effects from deforestation and selective log- INPE (2017b). Projeto TerraClass. Instituto Nacional de Pesquisas Espaciaishttp://www.
ging in the Brazilian Amazon. Biological Conservation, 141(7), 1745–1757. https:// inpe.br/cra/projetos_pesquisas/dados_terraclass.php, Accessed date: 6 August 2018.
doi.org/10.1016/j.biocon.2008.04.024. ISA (2015). Unidades de conservação na amazônia brasileira: Pressões e ameaças 2015.
Cabral, A. I. R., & Costa, F. L. (2017). Land cover changes and landscape patterns dy- Instituto Socioambientalhttps://www.socioambiental.org/sites/blog.socioambiental.
namics in Senegal and Guinea Bissau borderland. Applied Geography, 82, 115–128. org/files/nsa/arquivos/mapa_09set.pdf, Accessed date: 6 August 2018.
https://doi.org/10.1016/j.apgeog.2017.03.010. ISA (2017a). UCs Federais e Estaduais por categoria e estado na Amazônia Legal. https://uc.
Carreiras, J. M. B., Pereira, J. M. C., Campagnolo, M. L., & Shimabukuro, Y. E. (2006). socioambiental.org/c%C3%B4mputos/amaz%C3%B4nia-legal/grupos-e-categorias,
Assessing the extent of agriculture/pasture and secondary succession forest in the Accessed date: 6 August 2018.
Brazilian Legal Amazon using SPOT VEGETATION data. Remote Sensing of ISA (2017b). Terras Indígenas e Unidades de Conservação na Amazônia Legal Brasileira-
Environment, 101(3), 283–298. https://doi.org/10.1016/j.rse.2005.12.017. Janeiro 2017. Instituto Socioambientalhttps://www.socioambiental.org/pt-br/
Carvalho, G. O., Nepstad, D., McGrath, D., Diaz, M., del, C. V., Santilli, M., et al. (2002). mapas/terras-indigenas-e-unidades-de-conservacao-na-amazonia-legal-brasileira-
Frontier expansion in the Amazon: Balancing development and sustainability. janeiro-2017, Accessed date: 6 August 2018.
Environment: Science and Policy for Sustainable Development, 44(3), 34–44. https://doi. Joppa, L. N., Loarie, S. R., & Pimm, S. L. (2008). On the protection of “protected areas”.
org/10.1080/00139150209605606. PNAS, 105(18), 6673–6678. https://doi.org/10.1073/pnas.0802471105.
Cochrane, M. A., & Laurance, W. F. (2002). Fire as a large-scale edge effect in Amazonian Kaiser, H. F. (1960). The application of electronic computers to factor analysis.
forests. Journal of Tropical Ecology, 18(3), 311–325. https://doi.org/10.1017/ Educational and Psychological Measurement, XX(1), 141–151. https://doi.org/10.
S0266467402002237. 1177/001316446002000116.
Cochrane, M. A., & Laurance, W. F. (2008). Synergisms among fire, land use, and climate Kirby, K. R., Laurance, W. F., Albernaz, A. K., Schroth, G., Fearnside, P. M., Bergen, S.,
change in the Amazon. Ambio, 37(7–8), 522–527. https://doi.org/10.1579/0044- et al. (2006). The future of deforestation in the Brazilian Amazon. Futures, 38(4),
7447-37.7.522. 432–453. https://doi.org/10.1016/j.futures.2005.07.011.

114
A.I.R. Cabral et al. Applied Geography 100 (2018) 101–115

Lang, S., Klug, H., Langanke, T., & Tiede, D. (2004). Vector-based landscape analysis tools org/10.1098/rstb.2007.0036.
(extension for ArcGIS 10), version 2.0 beta: V-LATE 2.0 beta. Centre for Geoinformatics, Nicolle, S., & Leroy, M. (2017). Advocacy coalitions and protected areas creation process:
Z-GIS, University of Salzburg. Case study in the Amazon. Journal of Environmental Management, 198(1), 99–109.
Laurance, W. F. (2007). Switch to corn promotes Amazon deforestation. Science, https://doi.org/10.1016/j.jenvman.2017.04.035.
318(5857), 1721. https://doi.org/10.1126/science.318.5857.1721b. Nilsson, M. S. T., & Fearnside, P. M. (2011). Yanomami mobility and its effects on the
Lee, T. M., Sodhi, N. S., & Prawiradilaga, D. M. (2007). The importance of protected areas forest landscape. Human Ecology, 39(3), 235–256. https://doi.org/10.1007/s10745-
for the forest and endemic avifauna of Sulawesi (Indonesia). Ecological Applications, 011-9400-4.
17(6), 1727–1741. https://doi.org/10.1890/06-1256.1. Nobre, C. A., Sampaio, G., Borma, L. S., Castilla-Rubio, J. C., Silva, J. S., & Cardoso, M.
Linke, J., Fortin, M.-J., Courtenay, S., & Cormier, R. (2017). High resolution global maps (2016). Land use and climate change risks in the Amazon and the need of a novel
of 21st-century annual forest loss: Independent accuracy assessment and application sustainable development paradigm. PNAS, 113(39), 10759–10768. https://doi.org/
in a temperate forest region of Atlantic Canada. Remote Sensing of Environment, 188, 10.1073/pnas.1605516113.
164–176. https://doi.org/10.1016/j.rse.2016.10.040. Nogueira, E. M., Yanai, A. M., de Vasconcelos, S. S., Graça, P. M. L.de A., & Fearnside, P.
Lovejoy, T. E., & Nobre, C. A. (2018). Amazon tipping point. Science Advances, 4(2), M. (2018). Carbon stocks and losses to deforestation in protected areas in Brazilian
https://doi.org/10.1126/sciadv.aat2340 eaat2340. Amazonia. Regional Environmental Changes, 18(1), 261–270. https://doi.org/10.
Malhi, Y., Roberts, J. T., Betts, R. A., Killeen, T. J., Li, W., & Nobre, C. A. (2007). Climate 1007/s10113-017-1198-1.
change, deforestation and the fate of the Amazon. Scienceexpress. https://doi.org/10. Numata, I., & Cochrane, M. A. (2012). Forest fragmentation and its potential implications
1126/science.1146961 1 pp. in the Brazilian Amazon between 2001 and 2010. Open Journal of Forestry, 2(4),
Mascarenhas, F. S., Silva, S. S., & Brown, I. F. (2017). Dinâmica de Incêndios Florestais na 265–271. https://doi.org/10.4236/ojf.2012.24033.
reserva extractivista Chico Mendes - Acre. Anais do XVIII Simpósio Brasileiro de Pan, Y., Birdsey, R. A., Fang, J., Houghton, R., Kauppi, P. E., Kurz, W. A., et al. (2011). A
Sensoriamente Remoto, XVIII SBSR (pp. 6602–6608). 978-85-17-00088-1. large and persistence carbon sink in the world's forests. Science, 333(6045), 988–993.
Matricardi, E. A. T., Skole, D. L., Pedlowski, M. A., & Chomentowski, W. (2013). https://doi.org/10.1126/science.1201609.
Assessment of forest disturbances by selective logging and forest fires in the Brazilian Rempel, R. S., Kaukinen, D., & Carr, A. P. (2012). Patch analyst and patch grid. Ontario
Amazon using Landsat data. International Journal of Remote Sensing, 34(4), ministry of natural resources. Thunder Bay, Ontario: Center for Northern Forest
1057–1086. https://doi.org/10.1080/01431161.2012.717182. Ecosystem Research.
Maus, V. (2014). Satellite time series analysis for land use/cover change detection. IIASA Rossoni, L., Engelbert, R., & Bellegard, N. L. (2016). Normal science and its tools:
Interim Report. Laxenburg, Austria, IR-14-017 http://pure.iiasa.ac.at/11251/, Reviewing the effects of exploratory factor analysis in management. Revista de
Accessed date: 6 August 2018. Administração, 51(2), 198–211. https://doi.org/10.5700/rausp1234.
May, P. H., Gebara, M. F., de Barcellos, L. M., Rizek, M. B., & Milikan, B. (2016). The Saito, C. H., & Azevedo, A. A. (2017). Organic Intellectuals: Legitimizing agribusiness
context of REDD+ in Brazil: Drivers, actors and institutions (3rd ed.). Bogor, Indonesia: production in Brazil. International Gramsci Journal, 2(2), 107–132. http://ro.uow.edu.
CIFOR Occasional paper 160. au/gramsci/vol2/iss2/8, Accessed date: 6 August 2018.
McGarigal, K., Cushman, S. A., Neel, M. C., & Ene, E. (2002). FRAGSTATS: Spatial pattern Sasaki, N., & Putz, F. E. (2009). Critical need for new definitions of “forest” and “forest
analysis program for categorical maps. University of Massachusettshttp://www.umass. degradation” in global climate change agreements. Conservation Letters, 2(5),
edu/landeco/research/fragstats/fragstats.html. 226–232. https://doi.org/10.1111/j.1755-263X.2009.00067.x.
MMA (2011). Strategies to reduce deforestation in Brazil: From controlling illegal deforestation Skole, D., & Tucker, C. (1993). Tropical deforestation and habitat fragmentation in the
to the challenge of sustainable production in the country's forests and savannas. Ministério Amazon: Satellite data from 1978 to 1988. Science, 260(5116), 1905–1910. https://
do Meio Ambiente. http://www.stapgef.org/sites/default/files/stap/wp-content/ doi.org/10.1126/science.260.5116.1905.
uploads/2013/09/Strategies-to-Reduce-Deforestation-in-Brazil.pdf, Accessed date: 6 SNUC (2011). Sistema Nacional de Unidades de Conservação da Natureza (SNUC) e Plano
August 2018. Estratégico Nacional de Áreas Protegidas (PNAP). Ministério do Meio Ambiente. Brasília:
MMA (2016). Planos de ação para a prevenção e o controle do desmatamento. Documento Secretaria de Biodiversidade e Florestas, Departamento de Áreas Protegidas. http://
base: Contexto e análises. Versão preliminar aprovada pelo GPTI. http://www.mma.gov. www.mma.gov.br/images/arquivos/areas_protegidas/snuc/Livro%20SNUC%
br/images/arquivo/80120/PPCDAm%20e%20PPCerrado%20-%20Encarte 20PNAP.pdf, Accessed date: 6 August 2018.
%20Principal%20-%20GPTI%20_%20p%20site.pdf, Accessed date: 6 August 2018. Soares-Filho, B., Moutinho, P., Nepstad, D., Anderson, A., Rodrigues, H., Garcia, R., et al.
Molina, J. R., Martín, A., Drake, F., Martín, L. M., & Herrera, M. A. (2015). Fragmentation (2010). Role of Brazilian Amazon protected areas in climate change mitigation.
of Araucaria araucana forests in Chile: Quantification and correlation with structural PNAS, 107(24), 10821–10826. https://doi.org/10.1073/pnas.0913048107.
variables. iForest Biogeosciences and Forestry, 9, 244–252. https://doi.org/10.3832/ Tapia-Armijos, M. F., Homeier, J., Espinosa, C. I., Leuschner, C., & de La Cruz, M. (2015).
ifor1399-008. Deforestation and forest fragmentation in South Ecuador since the 1970s-Losing a
Moran, E. F. (1993). Deforestation and land use in the brazilian Amazon. Human Ecology, hotspot of biodiversity. PloS One, 10(9), https://doi.org/10.1371/journal.pone.
21(1), 21 pp. www.jstor.org/stable/4603072. 0133701 18 pp.
Morton, D. C., & DeFries, R. S. (2005). Rapid assessment of annual deforestation in the Tian, Y., Jim, C. Y., Tao, Y., & Shi, T. (2011). Landscape ecological assessment of green
Brazilian Amazon using MODIS data. Earth Interactions, 9(8), 21 pp. https://doi.org/ fragmentation in Hong Kong. Urban Forestry & Urban Greening, 10(2), 79–86. https://
10.1175/EI139.1. doi.org/10.1016/j.ufug.2010.11.002.
Nepstad, D., McGrath, D., Stickler, C., Alencar, A., Azevedo, A., Swete, B., et al. (2014). Tyukavina, A., Hansen, M. C., Potapov, P. V., Stehman, S. V., Smith-Rodriguez, K., Okpa,
Slowing Amazon deforestation through public policy and interventions in beef and C., et al. (2017). Types and rates of forest disturbance in brazilian legal Amazon,
soy supply chains. Science, 344(6188), 1118–1123. https://doi.org/10.1126/science. 2000-2013. Science Advances, 3(4), https://doi.org/10.1126/sciadv.1601047 15 pp.
1248525. Vedovato, L. B., Fonseca, M. G., Arai, E., Anderson, L. O., & Aragão, L. E. O. C. (2016).
Nepstad, D. C., Stickler, C. M., & Almeida, O. T. (2006). Globalization of the Amazon soy The extent of 2014 forest fragmentation in the Brazilian Amazon. Regional
and beef industries: Opportunities for conservation. Conservation Biology, 20(6), Environmental Changes, 16(8), 2485–2490. https://doi.org/10.1007/s10113-016-
1595–1603. https://doi.org/10.1111/j.1523-1739.2006.00510.x. 1067-3.
Nepstad, D. C., Stickler, C. M., Soares-Filho, B., & Merry, F. (2008). Interactions among Walker, R., Moore, N. J., Arima, E., Perz, S., Simmons, C., Caldas, M., et al. (2009).
Amazon land use, forests and climate: Prospects for a near-term forest tipping point. Protecting the Amazon with protected areas. PNAS, 106(26), 10582–10586. www.
Philosophical Transactions of The Royal Society B, 363(1498), 1737–1746. https://doi. pnas.orgcgidoi10.1073pnas.0806059106.

115