Policy Paper Government

Analyzing the Spatial Dynamics of Deforestation

in Brazilian Amazon

Daniel de Alencastro Bouchardet1, Alexandre Alves Porsse2, Romano Timofeiczyk Junior1

1 Forest and Wood Science Centre – Forest Engineering, Federal University of Parana, Curitiba,

Paran�a, Brazil, 2Department of Economics, Federal University of Parana, Curitiba, Paran�a, Brazil

Historically, development in Brazilian Amazon was promoted by permits of deforestation

under soft territory control or supervision. However, due to the importance of this biome

for biodiversity and ecosystem balance in a global perspective, Brazilian’s government

has tightened deforestation control. This work investigates the spatial pattern of defores-

tation in a cross-section and time perspective using global and local spatial data analysis.

Global results indicate the existence of high spatial correlation and that deforestation

holds concentrated across space, despite the efficacy of policy mechanisms adopted for

controlling and reducing the level of deforestation in Legal Amazon. Furthermore, local

results support the hypothesis of high spillover effects. Considering the spatial analysis

results, we highlight some implications for policy design aiming deforestation control.

Introduction

Deforestation in Brazilian Amazon has been discussed in different areas with many

approaches, mainly because of Amazon importance for biodiversity and ecosystem balance in

a global perspective (Fearnside 2008; Nepstad et al. 2009; Souza, Miziara, and Junior 2013).

The significance of rainforest has made Brazilian government take explicit political actions

since 2004 to control deforestation in the Legal Amazon Region (May, Millikan, and Gebara

2011). Many papers have dedicated attention for modeling the deforestation causes and some

studies took into account an explicit spatial approach to investigate the spatial pattern of defor-

estation process in Brazilian Amazon (Lorena and Lambim, 2007; Robalino and Pffaf, 2012;

Rosa, Souza, and Ewers 2012; Arima et al., 2007; Hargrave and Kis-Katos 2013; Rosa et al.,

2013).

According to Haining (2003), there are four types of spatial processes that operate in the

geographic space: diffusion, dispersion, exchange and rate transfer, and interaction. The diffu-

sion process occurs when some attribute is acquired by a population and, at some point in time,

it is possible to identify those individuals (or areas) that own the attribute. Contrary to

Correspondence: Daniel de Alencastro Bouchardet, UFPR - CIFLOMA - ENGENHARIA FLORES-

TAL - Av. Pref. Loth�ario Meissner, 632, CEP: 80210-170 - Jardim Botânico - Curitiba - Paran�a - Bra- sil - Phone: 155 41 3360-4206, FAX: 155 41 3360-4206 / 4231. e-mail: [email protected].

Submitted: May 26, 2015. Revised version accepted: March 21, 2016.

doi: 10.1111/gean.12105 23 VC 2016 The Ohio State University

Geographical Analysis (2017) 49, 23–35

diffusion, when an attribute spreads, the dispersal process consists of population movement.

The third process, exchange and rate transfer, refers to expenditure among regions and products

flow. The last one—interaction process—takes place when the outcome of one location influen-

ces and is influenced by the outcome of another. In light of this, literature concerning deforesta-

tion dynamics points out evidences for the existence of spatial process in deforestation

(Aguiar, Câmara, and Escada 2007; Robalino and Pfaff 2012).

The competition between forest land and alternative land uses is one of the main drivers of

deforestation, conditioned by both opportunity costs and net benefit maximization (Barbier and

Burgess 1997; Barbier, Burgess, and Grainger 2010). From another point of view, Piazza and

Roy (2015) characterize the economic and ecological conditions under which deforestation

may occur, considering the relationship of benefits brought by standing forests or by alternative

uses of land.

Fearnside (2008), assuming that the property definition of public land is weak, points out

three phases for the land transaction process. During the first phase, colonists and settlers clear

forest areas to determine ownership. Throughout the second phase, ranchers acquire the defor-

ested lands and decide on which use based on products and land prices. The third phase is char-

acterized by the transaction of land property from ranchers to capitalized farmers. As a

complement, Souza, Miziara, and Junior (2013) show evidences of higher deforestation in

areas with higher density of farmlands (private properties).

In summary, we have empirical reasons to believe that the deforestation mechanism sus-

tains a spatial process. As an example, the description in Fearnside (2008) shows evidence of

diffusion process. Further, the works of Robalino and Pfaff (2012) and Hargrave and Kis-Katos

(2013) showed significant spatial coefficients for deforestation, both covering rainforest areas.

However, the linkage between the spatial pattern of deforestation and policies build for control-

ling deforestation has been little explored in the empirical literature.

This study aims to investigate the spatial dynamic of deforestation process in Brazilian

Amazon in the context of the policies adopted over the last decade aiming to control or reduce

deforestation. First, we present a brief description on the deforestation policies built by Brazil-

ian authorities followed by a description of the database, variables, and the spatial techniques

used in this analysis. The results provide information about the dynamics of global and local

spatial dependency of deforestation. Finally, the implications of this spatial analysis for defor-

estation policies are discussed.

Overview of deforestation policy in Brazilian Amazon

Historically, the development in the Brazilian rainforest occurred based on permissions for

conversion of forestry land into agriculture or pasture, tax incentives for agriculture develop-

ment, and forest-selective cut (Dennis, van Riper, and Wood 2011). The National Institute of

Spatial Research (INPE) has published data on annual deforested area in Brazilian Amazon

since 1988, which allows monitoring the evolution of deforestation process. Since 2004, the

Brazilian government has announced legal mechanisms for controlling deforestation (May,

Millikan, and Gebara 2011) and deforested area in Legal Amazon has been decreasing annu-

ally, except for an increase in 2008 (11%) and another in 2013 (29%) (INPE 2015). Based on

policy instruments, we highlight 2004 and 2008 as key years for deforestation reduction, the

same considered by Assunç~ao, Gandour, and Rocha (2012).

Geographical Analysis

24

In 2004, the Brazilian government launched a national plan, known as PPCDAm, 1

for pre-

venting and controlling deforestation in Brazilian Amazon. This plan puts responsibilities to

federal, state, and municipal governments as well as private agents. Concisely, the main strat-

egies of PPCDAm are based on planning land use, monitoring and controlling deforestation,

and favoring sustainable production. Owing to changes in deforestation dynamics, PPCDAm

has passed through three phases since 2004. The first phase (2004–2008) focused on planning

land use by creating conservation unities (250 thousand km 2 ) and indigenous territories (100

thousand km 2 ). The second phase initiated in 2009 and lasted till 2011. Within this period, the

DETER 2

project, which provides daily information about deforestation in favor of better super-

vising, permitted an integrated action between the IBAMA, 3

federal police, highway federal

police, National Force, and Brazilian army for command and control mechanisms. The third

phase has been planned for 2012–2015 and aims to expand sustainable production (Minist�erio

do Meio Ambiente 2013).

Moreover, the presidential decree no. 6321, approved in 2007, assigns to the Ministry of

the Environment the responsibility of formulating an annual list containing risky municipal-

ities, based on municipal deforestation indicators, which should be monitored closely. The

resolution of National Monetary Council no. 3545, established in 2008, restrains properties

without proof of environmental regularity from receiving financial aid to invest in agriculture

or cattle. Also in 2008, by approving the National Law of Climate Change, the Brazilian gov-

ernment committed to reduce its projected emissions of greenhouse gases by 36.1% or 38.9%

until 2022. For this purpose, two funds that finance environmental projects were created: the

Climate Fund and the Amazon Fund. As its creation, Amazon Fund has financed 28 projects

statewide and 37 projects that focus on specific municipalities.

The policies implemented until 2006 were horizontal, affecting all economic agents in

Brazilian Amazon area, and proved successful for achieving a systematic reduction in the

deforested total area, despite some years of transitory inflection in the declining tendency. This

reduction happened mainly because the decrease in the frequency of large deforested areas

(Rosa, Souza, and Ewers 2012; Godar et al. 2014). The new policy mechanisms has raised the

importance of vertical actions, putting in evidence the need for incorporating spatial analysis

into the formulation and execution of deforestation policies.

Data and methods

The Brazilian government established Legal Amazon in 1959 covering approximately 5 mil-

lion hectares across nine Brazilian’s states: Acre, Amap�a, Amazonas, Mato Grosso, Rondônia,

Roraima, Par�a, Tocantins, and most of Maranh~ao.4 For the last decades, the main pressure for

converting forestry areas has concentrated the in arc of deforestation, which extends from

southeast of Maranh~ao, passing by Tocantins, Par�a, Mato Grosso, and Rondônia, and finish in

southeast of Acre. Deforested areas are mainly used for cattle production (May, Millikan, and

Gebara 2011).

The Brazilian INPE provides annual deforested area by municipality estimated using satel-

lite images. Because of weather conditions, the deforestation rate is calculated as the difference

between deforested areas in July of year t and august of year t 2 1. There is no differentiation

between legal and illegal deforestation in the satellite images data.

We considered the period between 2002 and 2013. Although there is data available for pre-

vious year, they should not be compared with 2002 because they are aggregated in the data

Daniel de Alencastro Bouchardet et al. Deforestation in Brazilian Amazon

25

available for 2001, the same situation described by Rosa, Souza, and Ewers (2012). The total

sample covers 760 municipalities. However, from these, we removed the sample municipal-

ities, which are covered with tropical savanna (Cerrado) 5

and municipalities that did not pres-

ent any forest area after 2002. Finally, we ended up with 686 municipalities, which were used

for applying the exploratory analysis of data techniques.

Variables

Searching to assess the spatial dynamics of deforestation, we used three variables. The first one

represents the annual deforested area (yit) measured in squared kilometers, commonly used for

evaluating the level of deforestation and for defining the arc of deforestation. In addition, we

use two percentage rates for deforestation, which allow assessing the intensity of the deforesta-

tion process over space. These rates were calculated as follows:

uit5 yit Si �100 (1)

where yit is the annual deforested area and si is the total municipal area, both in squared kilo-

meters, and uit provides the annual rate of deforestation. The second rate is determined as:

fit5

PT t51 yit Si

�100 (2)

where PT

t51 yit is the cumulative deforested area between t and T for each municipality and

thus fit provide the annual cumulative rate of deforestation. The motivation for using these three variables is to capture different aspects to discuss pol-

icy implications. First, the annual deforested area indicates regions that most contribute to

absolute deforestation. Alternatively, the rates indicate regions that suffered higher

Figure 1. Brazilian Amazon and the arc of deforestation.

Geographical Analysis

26

environmental degradation—in forest area conversion terms. Moreover, when compared with

the first rate [equation (1)], the second one [equation (2)] tries to capture a maximizing behav-

ior of alternative land uses. We expect that this rate presents a positive growth rate and, as

deforestation occurs and forest areas become scarcer, the growth rate decreases and the rate

value nears its maximum, that is, whether there is no more forests or the only forest areas

remaining must be protected by law and any deforestation is illegal.

Methods

The spatial dynamics of deforestation process is evaluated using the global and local measures

of spatial dependence. First, we calculate the Moran’s I for each variable, described as follows:

K5

Pn i51

Pn j51 wij yi2�yð Þ yj2�y

� � Pn

j51 yi2�yð Þ 2

(3)

where y is the variable of interest in deviations from the mean and wij represents the spatial

weights. Moran’s I informs about the existence and degree of spatial concentration (Almeida

2012, p. 106). Hence, if calculated by each year, it can provide dynamic information at least in

the global perspective.

The local dynamic is evaluated by using the Local Indicator of Spatial Association (LISA)

statistics, which represents the local decomposition of Moran’s I. Following Anselin (1995),

LISA statistics is represented as follows:

L ¼ zi Xj

j¼1 wijzj (4)

where zi and zj are the variables of interest in deviation from the mean. As is well known, by

incorporating neighboring behavior, LISA statistics allow identifying four types of clusters to

the deforestation process: high-high (HH), low-low (LL), high-low (HL), and low-high (LH).

HH and LL clusters occur when one region with high (low) value of the variable is surrounded

by regions with high (low) values. Conversely, HL and LH clusters occur when regions with

low deforestation are surrounded by polygons with high deforestation (LH) and regions with

high deforestation are surrounded by polygons with low deforestation (HL), respectively.

Applying this technique on the variables for each year allows investigating the extent to which

the spatial pattern of deforestation in terms of homogeneity and heterogeneity remains constant

or changes over time.

Spatial weight matrix

The selection of W (spatial weight matrix) can be broad and is critical for spatial analysis,

varying among a contingency matrix, distance-based or k-nearest neighborhoods. As described

previously, the original sample provided by INPE covers 760 municipalities and was reduced to

686 municipalities. Such an aspect of the database implies that some locations became “islands”

in the sense that the municipality does not present any physical border with its neighbors. Given

this condition, we employed the procedure used by Carvalho and Almeida (2010) for choosing

W. The structure of W was defined based on a k-nearest matrix with three neighborhoods, which

maximized the Moran’s I statistic after testing for higher-order k neighborhoods.

Daniel de Alencastro Bouchardet et al. Deforestation in Brazilian Amazon

27

Results and discussion

Variables

Figure 2 reports the mean and standard deviation of each variable calculated based on the

municipalities data. The systematic reduction observed in the mean and standard deviation for

variables y and u shows the efficacy of horizontal deforestation policies and suggests conver- gence dynamic in deforestation among municipalities. As expected, f presents a concave behavior suggesting that the cumulative deforestation rate could achieve a maximum value.

Global analysis

Moran’s I (K) results show the presence of spatial correlation for the three variables and the degree of spatial correlation is higher for the percentage rates when compared with the defor-

ested area (Fig. 3). Initially, the concentration of locations that most contribute to aggregate

deforestation, represented by y, presents the same behavior of average deforestation by munici-

pality with a peak in 2003 and with a second in 2008. However, rather than a decrease after

2008, K value has increased until 2013 (K 5 0.385), that is, regions that most contribute to aggregate deforestation in Legal Amazon are more concentrated in recent years, with values

similar to Moran’s I in 2002.

Results for deforestation controlled by municipal area (u) show that when average munici- pal deforestation presents peaks (2003 and 2008), the spatial correlation of u decreases when compared with the previous year. Considering that the average municipal area of our sample is

6,368 km 2

(standard deviation 13,512), with a maximum and minimum of 159,540 km 2

(municipality Altamira—PA) and 64 km 2

(municipality Raposa—MA), respectively, peaks in

total deforestation may occur together with an increase in deforestation of large municipalities.

Therefore, relative deforestation becomes less concentrated. Comparing 2002 and 2013,

Moran’s I for y increased 3% and for u decreased 9%. As f is a cumulative measure, the expected result was a positive variation. As seen in

Fig. 3, from 2004 to 2013, Moran’s I results are flat. The behavior of spatial correlation for f shows that municipalities with high deforested area were concentrated and, as the relative rate

of deforestation decreased in Brazilian Amazon since 2002 (see Fig. 2), less municipalities

detach from others because of relative high deforestation, maintaining K at the same level. Considering the spatial process described by Haining (2003, p. 21), the many levels of

deforestation within the municipalities through time cause the diffusion process. The exchange

of agriculture products between municipalities defines the exchange and rate transfer because

deforestation and cattle and agriculture production are correlated (Geist and Lambin, 2002).

Furthermore, because of this correlation, if there is an interaction between municipalities

Figure 2. Mean and standard deviation of the deforestation and the two rates.

Geographical Analysis

28

concerning the decision of agriculture production, them the interaction process have a part in

the observed spatial dependence.

Motel, Pirard, and Combes (2009) divide the causes of deforestation into two categories:

structural and political. Structural causes are related to local infrastructure and market factors.

Political causes are related to government instruments that result in higher deforestation (incen-

tives to agroindustry for example) or instruments that control deforestation. From another point

of view, Geist and Lambin (2002) divided the main causes of deforestation in proximate causes

(infrastructure and agriculture expansion, wood extraction) and underlying driving forces

(demographic, economic, technological, policy and institutional, and cultural factors). Proxi-

mate causes are related to human activity at the local level and underlying driving forces are

social process. The presence of spatial correlation reinforces that deforestation is related to

local characteristics, that is, there are near regions where deforestation occurs with higher fre-

quency. This conclusion is supported by the findings in Rosa et al. (2013) and Rosa et al.

(2015), that the deforestation presents a contagious process, mainly where there are roads near.

In recent years, the composition of deforested areas and of agents responsible for defores-

tation has changed. Since 2002–2009, the participation of small fragments in deforestation

(<50 ha) has increased from 25% to 70%, approximately (Rosa et al., 2012). Additionally, the

contribution of properties with more than 500 ha to deforestation decreased between 2004 and

2011 and policies adopted by the Brazilian government affected more effectively areas with

larger deforestation (Godar et al., 2014). However, the spatial correlation holds through the

entire period that we analyzed. However, the spatial correlation holds despite these changes

and the decrease in deforested area during the analyzed period. One implication is that policy

instruments are not acting to combat the contagious effect of deforestation and, consequently,

not fragmenting locations with high deforestation.

Local analysis

For LISA estimation, we used the same k-nearest matrix with three neighborhoods with 0.05%

of significance and results were submitted to a 10,000 permutation for robustness check. Defor-

estation clusters are revealed in Figs. 4–6. 6

Red polygons correspond to high-high (HH) clus-

ters and blue polygons to low-low (LL) clusters, and these are hot-spot polygons. Low-high

and high-low regions area are colored in purple and yellow, respectively. Gray polygons indi-

cate municipalities with no forest since 2002 and, consequently, are excluded from the

analysis.

Figure 3. Moran’s I statistics for municipal deforestation in Brazilian Legal Amazon.

Daniel de Alencastro Bouchardet et al. Deforestation in Brazilian Amazon

29

The low frequency of HL and LH clusters supports the argument that deforestation is a

border phenomenon and related with regional features, that is, there are regions more likely of

being deforested and once deforestation becomes infeasible, agents move toward nearest

Figure 4. LISA analysis for annual deforested area (y).

Figure 5. LISA analysis for annual relative deforestation index (u).

Geographical Analysis

30

regions to expand their activities. Rosa et al. (2013) found this contagious process for defores-

tation in Brazilian Amazon.

Considering the annual deforested area (Fig. 4), the location of low-low (LL) clusters

holds from south to northeast borders. From 2002 to 2013, the high-high (HH) cluster shifted

to north and no longer covers the north region of Mato Grosso. In 2002, the state of Amazonas

had only 1 municipality as an HH region, but in 2013, 6 municipalities were located in this

cluster. In Rondônia, the same pattern occurred: in 2002, there were two municipalities in HH

cluster, and in 2013, there were nine. Conversely, the state of Mato Grosso presented 29 munic-

ipalities in HH clusters and 8 municipalities remained as HH cluster at the end of the period.

Initially, there were 44 municipalities in high-high clusters and 121 in low-low clusters. In

2013, these numbers were 34 and 139, respectively.

The production of soy and maize in Mato Grosso has increased since 2002, mainly in the

center of the state according to the Municipal Agricultural Survey, elaborated by the Brazilian

Institute of Geography and Statistics (IBGE/PAM, 2015). Hence, the movement of the high-

high cluster towards north may be indicative of a new agriculture frontier. The relation between

deforestation and agriculture is also reported by Ewers, Lawrance, and Souza (2008). Despite

they did not find a casual-effect relationship between agriculture and deforestation, historically

the correlation between price and agriculture area is high and this relation may impact defores-

tation because of forest clearing for agriculture expansion.

The occurrence of HH cluster reduces when we control deforestation by municipal area

(Fig. 5). In 2002, HH clusters covered 43 municipalities and in 2013, this number dropped to

34. Similar to the annual deforested area results (Fig. 4), the area of the cluster in central Mato

Grosso decreases. Furthermore, the HH cluster in the border of Par�a and Maranh~ao defrag-

mented after 2009. The only HH cluster in 2013 was located in the border of the north of

Rondônia connected with the south of Amazonas.

Figure 6. LISA analysis for annual cumulative deforestation index (f).

Daniel de Alencastro Bouchardet et al. Deforestation in Brazilian Amazon

31

Lastly, Fig. 5 shows f results. Like global spatial analysis (Fig. 3), since 2004 the pattern of spatial distribution holds. From 2002 to 2004, the number of municipalities in-hot spots

increased: HH clusters had 43 municipalities in 2002 and 74 in 2004; LL clusters had 117

municipalities in 2002 and 129 in 2004.

Despite y is the most common indicator for deforestation in Legal Amazon, the percentage

rates allow to identify regions that suffered higher relative deforestation. Considering the varia-

tion on municipalities areas of Legal Amazon, a relative measure should be considered when

policies area aiming to reduce environmental degradation, and not only the reduction of

deforestation.

Policy implications

We refer to policy instruments that cover all Legal Amazon territory as “horizontal policies.”

For example, the second phase of PPCDAm was a horizontal mechanism being focused in

command and control mechanisms by enhancing deforestation monitoring. Another example is

projects financed by Amazon Fund that cover all states of Legal Amazon. From 2009 to 2013,

the Amazon Fund financed 63 projects in Legal Amazon region, among which 29 are statewide

(Fundo Amazônia 2015). The creation of Conservation Unities can also be considered as a hor-

izontal instrument, except when the assumptions for creation are based on regional aspects, as

for creating green barriers for deforestation.

The expected result of effective horizontal policy instruments is the decrease of annual

deforested area in Legal Amazon. Data published by INPE show evidences of effective results

due to lower levels in historical deforested area after 2004. Moreover, Hargrave and Kis-Katos

(2013) show that fines disobeying environmental laws intensity have a significant effect on

deforestation reduction. However, our results show that spatial correlation holds despite the

reduction in total deforested area, as seen by Moran’s I results.

From another perspective, “vertical policy” instruments are concerned about specific

regions. The presidential decree no. 6321 and the resolution no. 3545 of Monetary Council

affect municipalities, which are considered as key locations for decreasing deforestation. Also,

34 projects financed by Amazon Fund cover specific municipalities. The expected effect of ver-

tical policy is the decrease of deforestation in municipalities where deforestation highlights. In

this context, the results of Assunç~ao et al. (2013) show that the credit constraint policy (resolu-

tion no. 3545) led to a reduction in deforestation levels.

The spatial interpretation of vertical policy effectiveness would be less frequency or

smaller area of high-high clusters. LISA results of municipal deforestation controlled by

municipal area (u) show that HH clusters are decreasing and we could interpret this fact as an evidence of the vertical policy efficacy. However, the results of LISA estimated for annual

deforested area show persistence of concentrated regions with high deforestation. We point out

two hypotheses for these results. The first one is that deforestation in municipalities that are

perceived as critical and suffer deforestation control returns to its original relative levels after

the implementation of policy instruments ceases. The second is that when deforestation in criti-

cal municipalities is controlled, deforestation in other regions grows because the contagious

effect, as seen by the shift of HH clusters to north when considering the annual deforested area.

This transition of deforestation to new areas can be associated with failures in the Brazilian

land legal framework which do not provide adequate protection to landholders. Thus deforesta-

tion is a mechanism for legitimate land ownership (Araujo et al. 2009).

Geographical Analysis

32

Together with deforestation regulation, initiatives to recover environmental degradation

are important for Amazon ecosystem balance. Since 2009, Amazon Fund has financed 20 proj-

ects that aimed to recover deforested areas. LISA maps for f, the cumulative rate, show regions that historically detach because of higher deforestation and, consequently, suffered higher envi-

ronmental degradation. Therefore, LISA maps for f indicate regions that mostly demand envi- ronmental degradation recovery actions.

From another perspective, actions that increase the value of forest land and sustainable

production should contribute to decrease in deforestation, assuming that these actions reduce

the attractiveness of alternative land uses (Barbier, Burgess, and Grainger 2010). The objec-

tives of the third phase of PPCDAm are aligned with this strategy because of focus in promo-

tion of sustainable production. Considering the rate f, municipalities in high-high clusters suffered higher relative deforestation and are more likely to present deforestation in newly

grown forest areas because of infrastructure or territory composition and are potential targets

for policy instruments for sustainable production and increase in forest value.

Final remarks

Our results show that deforestation is spatially correlated as pointed by others works that inves-

tigated this matter (Lorena and Lambim, 2007; Robalino and Pffaf, 2012; Rosa, Souza, and

Ewers, 2012; Arima et al., 2007; Hargrave and Kis-Katos 2013; Rosa et al., 2013). However,

using a spatial exploratory data analysis we were able to bring insights about the spatial pattern

of deforestation and the policy instruments used to control this process.

Despite the effectiveness of PPCDAm in reducing aggregated level of deforestation

(Assunç~ao, Gandour, and Rocha 2012), our analysis suggests that this policy has been not

much effective to change the spatial distribution of deforestation and for reducing the defores-

tation spillover effects. This put in evidence the need for enhancing spatial aspects into the

framework of policies for controlling deforestation in the Brazilian Amazon. Additionally, the

maps provided by the local analysis carried out in this study allow to identify the regions more

sensitive to deforestation which should be the focus of local policies.

As our analysis concerns only about the linkage between the spatial pattern of deforesta-

tion and the deforestation policies implemented in the last decades, we have not explored

which policy actions could be more effective for controlling deforestation in a spatial perspec-

tive. This is an important issue to be explored by further research.

Notes

1 Action plan for the prevention and control of deforestation in the legal Amazon.

2 Real time system for detection of deforestation.

3 Brazilian institute for the environment and renewable natural resources.

4 To the west of meridian 448 west.

5 Deforestation in Cerrado areas is not reported by INPE.

6 The results are presented to four subperiods to exhibit the dynamics of cluster configuration over the

entire period.

References

Aguiar, A. P. D., G. Câmara, and M. I. S. Escada. (2007). “Spatial Statistical Analysis of Land-Use Deter-

minants in the Brazilian Amazonia: Exploring Intra-Regional Heterogeneity.” Ecological modelling 209(2), 169–88.

Daniel de Alencastro Bouchardet et al. Deforestation in Brazilian Amazon

33

Almeida, E. (2012). Econometria Espacial Aplicada, 1st ed. Campinas, SP: Al�ınea. Anselin, L. (1995). “Local Indicators of Spatial Association—LISA.” Geographical analysis 27(2), 93–115. Araujo, C., C. A. Bonjean, J. L. Combes, P. C. Motel, E. J. Reis. (2009). “Property Rights and Deforesta-

tion in the Brazilian Amazon.” Ecological Economics 68(8), 2461–8. Arima, E. Y., C. S. Simmons, R. T. Walker, M. A. Cochrane. (2007). “Fire in the Brazilian Amazon: A

Spatially Explicit Model for Policy Impact Analysis.” Journal of Regional Science 47(3), 541–67. Assunç~ao, J., C. C. Gandour, and R. Rocha. (2012). “Deforestation Slowdown in the Legal Amazon: Pri-

ces or Policies.” Climate Policy Initiative Working Paper. Rio de Janeiro, RJ, Brazil. Assunç~ao, J., C. Gandour, R. Rocha, and R. Rocha. (2013). “Does Credit Affect Deforestation? Evidence

from a Rural Credit Policy in the Brazilian Amazon.” Climate Policy Initiative Working Paper. Rio de Janeiro, RJ, Brazil.

Barbier, E. B., and J. C. Burgess. (1997). “The Economics of Tropical Forest Land Use Options”. Land Economics 73(2), 174–95.

Barbier, E. B., J. C. Burgess, and A. Grainger. (2010). “The Forest Transition: Towards a More Compre-

hensive Theoretical Framework.” Land Use Policy 27(2), 98–107. Carvalho, T. S., and E. Almeida. (2010). “A Hip�otese da Curva de Kuznets Ambiental Global: Uma Per-

spectiva Econom�etrico-Espacial.” Estudos Econômicos 40(3), 587–615. Dennis, K., C. J. van Riper, and M. A. Wood. (2011). “Payments for Ecosystem Services as a Potential

Conservation Tool to Mitigate Deforestation in the Brazilian Amazon.” Applied Biodiversity Scien- ces Perspectives Series 2(1), 1–15.

Ewers, R. M., W. F. Laurance, C. M. Souza. (2008) “Temporal Fluctuations in Amazonian Deforestation

Rates.” Environmental Conservation 35(4), 303–10. Fearnside, P. M. (2008). “The Roles and Movements of Actors in the Deforestation of Brazilian Ama-

zonia.” Ecology and Society 13(1), 23. Fundo Amazônia (2015). Carteira de Projetos. Fundo Amazônia Database. Accessed May 2015 from

http://www.fundoamazonia.gov.br/FundoAmazonia/fam/site_pt/Esquerdo/Projetos_Apoiados/.

Geist, J. H., and E. F. Lambin (2002). “Proximate Causes and Underlying Forces of Tropical

Deforestation.” Bioscience 52(2), 143–50. Godar, J., A. G. Toby, E. J. Tizado, P. Pacheco. (2014). “Actor-specific Contributions to the Deforestation

Slowdown in the Brazilian Amazon.” PNAS 111(43), 15591–96. Haining. R. P. (2003). Spatial Data Analysis: Theory and Pratice, 1st ed. University of Cambridge, U.K. Hargrave, J. and K. Kis-Katos. (2013). “Economic Causes of Deforestation in the Brazilian Amazon: A

Panel Data Analysis for the 2000s.” Environmental and Resource Economics 54(4), 471–94. IBGE/PAM (2015). Pesquisa Agr�ıcola Municipal. IBGE Database. Accessed May 2015 from http://

www.sidra.ibge.gov.br/bda/tabela/listabl.asp?c51612&z5t&o53.

INPE (2015). “Projeto PRODES – Monitoramento da Floresta Amazônica Brasileira por Sat�elite.” INPE Database. Accessed January 2015 from http://www.dpi.inpe.br/prodesdigital/prodesmunici- pal.php

Lorena, R B., and E. F. Lambin. (2007). “Linking Spatial Patterns of Deforestation to Land Use Using

Satellite and Field Data.”IEEE International Geoscience and Remote Sensing Symposium, 2007. Flo- rian�opolis, SC, Brazil, (IGARSS 2007).

May, P. H., B. Millikan, and M. F. Gebara. (2011). “O Contexto de REDD1 no Brasil: Determinantes,

Atores e Instituiç~oes.” Publicaç~ao Ocasional 62. CIFOR, Bogor, Indon�esia. Minist�erio do Meio Ambiente (2013). “Plano De Aç~ao Para Prevenç~ao e Controle do Desmatamento na

Amazônia Legal (PPCDAm): 3 a

Fase (2012-2015) Pelo Uso Sustent�avel e Conservaç~ao da Floresta”. Minist�erio do Meio Ambiente. Accessed March 2015 from http://www.fundoamazonia.gov.br/Fun- doAmazonia/export/sites/default/site_pt/Galerias/Arquivos/Publicacoes/PPCDAm_3a_fase.pdf

Motel, P. C., R. Pirard, and J. L. Combes. (2009). “A Methodology to Estimate Impacts of Domestic Poli-

cies on Deforestation: Compensated Successful Efforts for “Avoided Deforestation” (Redd).” Eco- logical Economics 68(3), 680–91.

Nepstad, D., B. S. Soares-Filho, F. Merry, A. Lima, P. Moutinho, J. Carter, M. Bowman, A. Cattaneo, H.

Rodrigues, S. Schwartzman, D. G. McGrath, C. M. Stickler, R. Lubowski, P. Piris-Cabezas, S.

Rivero, A. Alencar, O. Almeida, and O. Stella. (2009). “The End of Deforestation in the Brazilian

Amazon.” Science 326(5958), 1350–1.

Geographical Analysis

34

Piazza, A., and S. Roy. (2015). “Deforestation and Optimal Management.” Journal of Economic Dynam- ics and Control 53, 15–27.

Robalino, J. A., and A. Pfaff. (2012). “Contagious Development: Neighbor Interactions in Deforestation.”

Journal of Development Economics 97(2), 427–36. Rosa, I. M. D., D. Purves, J. M. B. Carreiras, R. M. Ewers. (2015). “Modelling Land Cover Change in the

Brazilian Amazon: Temporal Changes in Drivers and Calibration Issues.” Regional Environmental Change 15, 123–37.

Rosa, I. M. D., D. Purves, C. Souza Jr., R. M. Ewers. (2013). “Predictive Modelling of Contagious Defor-

estation in the Brazilian Amazon.” PLoS ONE 8(10): e77231. doi:10.1371/journal.pone.0077231 Rosa, I. M. D., C. Souza Jr., R. M. Ewer. (2012). “Change in Size of Deforestation Patches in the Brazil-

ian Amazon.” Conservation Biology 26(5), 932–7. Souza, R. A., F. Miziara, and P. D. M. Junior. (2013). “Spatial Variation of Deforestation Rates in the

Brazilian Amazon: A Complex Theater for Agrarian Technology, Agrarian Structure and Gover-

nance by Surveillance.” Land Use Policy 30(1), 915–24.

Daniel de Alencastro Bouchardet et al. Deforestation in Brazilian Amazon

35

Copyright of Geographical Analysis is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.