This article was downloaded by: [University of Chicago Library]On: 28 April 2013, At: 08:23Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

International Journal ofRemote SensingPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/tres20

Space-time dynamics ofdeforestation in BrazilianAmazôniaD. S. AlvesVersion of record first published: 25 Nov 2010.

To cite this article: D. S. Alves (2002): Space-time dynamics of deforestation inBrazilian Amazônia, International Journal of Remote Sensing, 23:14, 2903-2908

To link to this article: http://dx.doi.org/10.1080/01431160110096791

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or makeany representation that the contents will be complete or accurate orup to date. The accuracy of any instructions, formulae, and drug dosesshould be independently verified with primary sources. The publishershall not be liable for any loss, actions, claims, proceedings, demand, orcosts or damages whatsoever or howsoever caused arising directly orindirectly in connection with or arising out of the use of this material.

int. j. remote sensing, 2002, vol. 23, no. 14, 2903–2908

Space–time dynamics of deforestation in Brazilian Amazonia

D. S. ALVES

Divisao de Processamento de Imagens, Instituto Nacional de PesquisasEspaciais, Av. dos Astronautas 1758, Sao Jose dos Campos, SP, CEP12227-010 , Brazil; e-mail: dalves@dpi.inpe.br

(Received 10 January 2001; in � nal form 3 August 2001 )

Abstract. Deforestation maps derived from Landsat Multi-Spectral Scanner(MSS) and Thematic Mapper (TM) imagery were used to analyse spatial patternsof deforestation in the 1970s and the 1991–1997 period in Brazilian LegalAmazonia. Nearly 90% of the deforestation has occurred within 100 km frommajor roads established under federal development programmes. Clearings largerthan 50 ha and 200 ha accounted, respectively, for 74% and 50% of the totaldeforestation in the 1991–1997 period. Results show that more intense deforesta-tion has been concentrated over some regions, leading to the continuous enlarge-ment of forest clearings and contributing to aggravate deforestation impacts insuch areas.

1. IntroductionThe deforested area of the Brazilian Legal Amazonia has increased from 10 mil-

lion hectares in the 1970s to nearly 59 M ha in 2000 (Tardin et al. 1980, InstitutoNacional de Pesquisas Espaciais (INPE) 2001). The fast advance of deforestationfollowed 1970s and 1980s Brazilian government policies that included the buildingof extensive road networks, federal government appropriation of a 100 km zone toeach side of major roads for colonization projects, and government-assisted migrationand farming (Schmink and Wood 1992, Becker 1997, Machado 1998). Forest destruc-tion and the rapid expansion of the agricultural frontier aVects ecosystems, land–atmosphere interactions and climate (Melillo 1994, Gash et al. 1996), and has causedland degradation, social tension and rapid, precarious urbanization (Schmink andWood 1992, Becker 1997, Machado 1998).

In this study, deforestation maps derived from Landsat Multi-Spectral Scanner(MSS) and Thematic Mapper (TM) imagery were used to analyse the spatial distribu-tion of deforestation, and to evaluate how forest clearings of diVerent sizes contributedto total deforestation. The aims are to locate areas of intense forest clearing and toinvestigate how deforestation has evolved there. This analysis can contribute to under-stand forest-clearing mechanisms and help to reduce and mitigate deforestationimpacts.

2. Data and methodsThis study was based on deforestation maps generated by manual interpretation

of forest clearings observed on Landsat MSS and TM images. TM-derived maps

Internationa l Journal of Remote SensingISSN 0143-1161 print/ISSN 1366-590 1 online © 2002 Taylor & Francis Ltd

http://www.tandf.co.uk/journalsDOI: 10.1080/01431160110096791

Dow

nloa

ded

by [

Uni

vers

ity o

f C

hica

go L

ibra

ry]

at 0

8:23

28

Apr

il 20

13

D. S. Alves2904

contained forest clearings observed on 1:250000 scale images from years 1991–1997(INPE 2001). MSS-derived maps were based on 1:500000 scale images and containeddeforested areas observed in the � rst and second halves of the 1970s (Tardin et al.

1980 ).The individual deforestation maps derived from the TM data were aggregated

into digital mosaics for which the distribution of deforestation and forest clearingsize were analysed. Annualized 1991–1997 average deforestation rates were basedon imagery acquisition dates, and forest clearing sizes were obtained from aggregated

maps corresponding to the 1991–1994, 1994–1997 and 1991–1997 periods. Areascovered by non-forest vegetation (savannahs) and by clouds in 1991 or in 1997 wereexcluded from analysis. MSS-derived maps originated one single digital map of theaccumulated deforestation by the late 1970s.

Both TM- and MSS-derived maps were intersected with buVer zones aroundmajor roads within the 100 km limits of land appropriation based on Machado(1998). The amount of deforestation was determined for � ve groups of roads(� gure 1(a)): (1) ‘West’, linking the cities of Cuiaba, to Porto Velho, Rio Branco, and

western Acre; (2) ‘East’, linking Belem and Sao Luis to Palmas do Tocantins, Barrado Garcas, Maraba, Redencao, Sao Jose do Xingu, Altamira, and Sao Felix doXingu; (3 ) ‘Central’, connecting Cuiaba to Santarem, with links to Alta Floresta,and north-west Mato Grosso; (4) ‘Transamazon’, de� ned here as the link from

Maraba to Altamira, Itaituba and Labrea; and (5) ‘Northwest’, linking Porto Velhoto Manaus and Boa Vista, continuing northward.

The maps were also intersected with a 1/4° cell grid to evaluate the geographicdistribution of deforestation. Furthermore, groups of adjacent 1/4° cells presenting

spatially auto-correlated deforestation were also determined based on the estimationof a local indicator of spatial auto-correlation (Anselin 1995) for the 1970s accumu-lated deforestation and the 1991–1997 annual average deforestation rate.

The relative contribution of forest clearings was determined for six diVerent sizecategories (å 50 ha, 51–100 ha, 101–200 ha, 201–500 ha, 501–1000 ha and >1000 ha)

for the 1991–1994, 1994–1997 and 1991–1997 periods. The amount of 1991–1997deforestation for 1/4° cells was also determined for the same size categories, and

iterative aggregation of the cells was used to determine regions for which more than50% of the 1991–1997 deforested area was due to clearings of some of the size

categories.

Figure 1. (a) Map of the west, east, central, transamazon and northwest roads, and LegalAmazonia States (Acre, Amapa, Amazonas, Maranhao, Mato Grosso, Para, Rondonia,Roraima and Tocantins). (b) Cumulative distributions for 1/4° cells calculated for the1991–1997 annual average deforestation rate and the fractions of the 1991–1997deforested area due to forest clearings smaller than 100 ha and larger than 200 ha. Celldeforestation rate decreases from left to right. (c) Map of the 1/4° cells that presentedthe highest and the lowest deforestation rates and accounted for, respectively, 95%and 5% of the total rate. (d) Map of the 1/4° cells that presented signi� cant spatialauto-correlation (SA-C) ( p<0.05) for the 1970s accumulated deforestation and for the1991–1997 deforestation rate. (e) Map of regions for which forest clearings å 50 ha,>50 ha, >100 ha, >200 ha, >500 ha and >1000 ha represented more than 50% ofthe 1991–1997 deforested area, for 1/4° cells that presented the highest deforestationrates and accounted for 95% of the total rate (c).

Dow

nloa

ded

by [

Uni

vers

ity o

f C

hica

go L

ibra

ry]

at 0

8:23

28

Apr

il 20

13

Remote Sensing L etters 2905

3. Results and discussionThe cumulative distribution of the 1991–1997 annual average deforestation rate

by 1/4° cells is shown in � gure 1(b). The highest deforestation rates were concentratedin relatively few cells, with 27.7% of the cells that presented the highest deforestationrates accumulating 95% of the total rate. Figure 1(c) shows that these cells weremostly con� ned to the 100 km zone around major roads from � gure 1(a).

In fact, areas within 100 km from major roads concentrated nearly 90% of the

(a) (b)

(c)

(e)

(d)

Dow

nloa

ded

by [

Uni

vers

ity o

f C

hica

go L

ibra

ry]

at 0

8:23

28

Apr

il 20

13

D. S. Alves2906

analysed deforestation, as detailed in table 1, and the majority of the cells thatpresented the highest deforestation rates and accumulated 95% of the total rate(� gure 1(c)). Moreover, areas within 100 km from major roads in the States ofMaranhao, Mato Grosso, eastern Para and Rondonia accounted for 80% of the1991–1997 deforestation rate. Deforestation presented several groups of contiguouscells where signi� cant spatial auto-correlation was found ( p<0.05) for both the1970s accumulated deforestation and the 1991–1997 deforestation rate (� gure 1(d )).Most of these groups of cells were located within the 100 km zone around majorroads, 1970s development zones (the so-called ‘Polos de Desenvolvimento’) inMaranhao, Mato Grosso, eastern Para and Rondonia and inside the limits of theGrande Carajas Programme (Machado 1998), at the same time as spatial auto-correlated cells in 1991–1997 appear to have expanded around those from the 1970s.This suggests the formation of extensive areas of intense, spatially auto-correlateddeforestation in these regions, where deforestation impacts can be intensi� ed. Onthe other hand, signi� cant spatial auto-correlation was not found in more remoteregions, particularly along most of the Transamazon and Northwest roads.

The fractions of deforestation due to forest clearings of diVerent sizes are presentedin table 2 for the 1991–1994, 1994–1997 and 1991–1997 periods. The 1991–1994 and1994–1997 periods showed relatively similar partitioning among size categories,despite a 28% increase in the annual average deforestation rate in the later period(INPE 2001). At the same time, forest clearings larger than 1000 ha increased theirrelative contribution in 1991–1997 in comparison with the two shorter periods, whileclearings smaller than 50 ha decreased their contribution. Such relative increase ofthe largest forest clearings is an eVect of the successive connection of cleared areas

Table 1. Fractions of deforestation in areas within 100 km from major roads from � gure 1(a).Fractions based on an annual average deforestation rate of 1.6×106 ha yÕ 1 in 1991–1997 and 9.8×106 ha deforested by the end of the 1970s. Total includes deforestationfrom all groups, after deduction of overlapping areas.

Fraction of the 1991–1997 annual average deforestation rate(fraction of 1970s accumulated deforestation, %)

Distance Roads Trans- North-(km) West East Central amazon west Total

25 10.3 (17.0) 20.0 (30.0) 17.7 (12.3) 3.8 (3.5) 1.2 (1.5) 47.4 (58.1)50 17.0 (21.4) 31.5 (42.5) 25.4 (17.2) 6.3 (4.4) 2.2 (1.8) 72.6 (75.7)75 21.1 (22.7) 37.5 (49.3) 29.7 (20.9) 8.4 (5.7) 2.9 (1.9) 86.2 (84.4)

100 24.4 (23.7) 39.4 (51.3) 32.9 (23.8) 10.7 (6.9) 3.6 (2.2) 92.4 (87.5)

Table 2. Deforestation by clearing size categories in the 1991–1994, 1994–1997 and1991–1997 periods.

Fraction of total deforestation in size category (%)

Period å 50 ha 51–100 ha 101–200 ha 201–500 ha 501–1000 ha >1000 ha

1991–1994 31.2 13.0 12.8 15.3 10.0 17.71994–1997 32.3 12.5 11.8 14.3 9.5 19.61991–1997 26.5 11.8 11.9 14.9 10.5 24.4

Dow

nloa

ded

by [

Uni

vers

ity o

f C

hica

go L

ibra

ry]

at 0

8:23

28

Apr

il 20

13

Remote Sensing L etters 2907

in regions of deforestation concentration that can be observed on time series of� ne-resolution imagery.

In general, larger forest clearings generally accounted for the largest fractions ofthe 1991–1997 deforestation in cells of higher deforestation rates. Here, this tendencyis shown for clearings >200 ha (� gure 1(b)), whose relative contribution decreasedas cell deforestation decreased. On the other hand, the cumulative distribution fromclearings å 100 ha increased as cell deforestation decreased. At the same time, therelative importance of diVerent clearing sizes varied from region to region as shownin � gure 1(e). In general, forest clearings larger than 100 or 200 ha contributed tomore than 50% of the 1991–1997 deforested area in most of Mato Grosso andeastern Para. More than 70% of the deforested area were due to clearings >50 hain regions originally occupied by small farmers in eastern Para and Rondonia(Schmink and Wood 1992, Machado 1998); on the other hand, clearings <50 harepresented 62% of the deforestation in areas of small farm colonization along theTransamazon road in central Para.

It would be diYcult to relate forest clearing size to farm size, because of diVerentland use practices, speculation and land tenure concentration occurring in theAmazon (Machado 1998). However, analysis of census data shows that most defor-estation occurs in medium (>100 ha) and large farms (>1000 ha) (Fearnside 1993),suggesting that the predominance of larger forest clearings might be explained bydeforestation in medium and large farms.

Overall, it can be concluded that concentration of deforestation around majorroads and some development zones have originated extensive regions of high, spa-tially auto-correlated deforestation, where forest clearings tended to merge after sometime creating enlarged deforested areas. It can be noted that most deforestation hasoccurred where roads oVer relatively eYcient access to more developed regions ofBrazil, suggesting a scenario of land use intensi� cation associated with increaseddeforestation in such regions.

The outcome of concentration of forest clearings in Brazilian Amazon is thedevelopment of landscapes mostly dominated by pastures, with increasing forest andhabitat fragmentation (Skole and Tucker 1993), elimination of obligatory foresttracts in farmland (Alves et al. 1999), and potential intensi� cation of other negativedeforestation impacts.

AcknowledgementsThe author wishes to thank G. Camara, C. Nobre and E. Novo for comments

on an earlier text, and R. C. Souza for the help in putting together some of thedatasets used in this study. This work is the result of research partially funded byCNPq, FAPESP, and FINEP/PPG7.

References

Alves, D. S., Pereira, J. L. G., Souza, C. L., Soares, J. V., and Yamaguchi, F., 1999,Characterizing landscape changes in Central Rondonia using TM imagery.International Journal of Remote Sensing, 20, 2877–2882.

Anselin, L., 1995, Local indicators of spatial association – LISA. Geographical Analysis,27, 93–115.

Becker, B. K., 1997, Amazonia, 5th edn (Sao Paulo: Atica).Fearnside, P. M., 1993, Deforestation in Brazilian Amazonia: the eVect of population and

land tenure. Ambio, 22, 537–545.

Dow

nloa

ded

by [

Uni

vers

ity o

f C

hica

go L

ibra

ry]

at 0

8:23

28

Apr

il 20

13

Remote Sensing L etters2908

Gash, J. H. C., Nobre, C. A., Roberts, J. M., and Victoria, R. L. (eds), 1996, AmazonianDeforestation and Climate (New York: John Wiley & Sons).

Instituto Nacional de Pesquisas Espaciais (INPE), 2001, Monitoring of the BrazilianAmazon forest by satellite 1999–2000, Separata. Instituto Nacional de PesquisasEspaciais, Brazil.

Machado, L. A., 1998, A fronteira agricola na Amazonia brasileira. In Geogra� a e MeioAmbiente no Brasil, edited by B. K. Becker, A. Christofoletti, F. R. Davidovich andP. R. Geiger, 2nd edn (Sao Paulo: Hucitec), pp. 181–217.

Melillo, J. M., 1994, Modelling land–atmosphere interactions: a short review. In Change inL and Use and L and Cover: A Global Perspective, edited by W. B. Meyer and B. L.Turner II, (Cambridge: Cambridge University Press), pp. 387–409.

Schmink, M., and Wood, C. H., 1992, Contested Frontiers in Amazonia (New York: ColumbiaUniversity Press).

Skole, D. L., and Tucker, C. J., 1993, Tropical deforestation and habitat fragmentation inthe Amazon: satellite data from 1978 to 1988. Science, 260, 1905–1910.

Tardin, A. T., Lee, D. C. L., Santos, R. J. R., Assis, O. R., Barbosa, M. P. S., Moreira,M. L., Pereira, M. T., Silva, D., and Santos Filho, C. P., 1980, Subprojeto desmata-mento: convenio IBDF/CNPq—INPE. INPE-1649-RPE/103, Instituto de PesquisasEspaciais, Brazil.

Dow

nloa

ded

by [

Uni

vers

ity o

f C

hica

go L

ibra

ry]

at 0

8:23

28

Apr

il 20

13