---

Physical and Human Dimensions of
Deforestation in Amazonia

In the Brazilian Amazon, regional trends are influenced by large-scale external forces but mediated by local conditions

D. L. Skole, W. H. Chomentowski, W. A. Salas, and A. D. Nobre

Understanding of tropical deforestation, an important aspect of global change, is inadequate for two reasons: a lack of accurate measurements of its rate, geographic extent, and spatial pattern and a lack of insight into its causes (Skole in press). It seems obvious that tropical deforestation is the consequence of a variety of interrelated social, economic, and environmental factors. Yet, interpretations of how these factors interact to stimulate deforestation vary widely. Some interpretations focus chiefly on population growth, whereas others regard institutions as the main determinant (Allen and Barnes 1985, Browder 1988, Bunker 1984b, Meyer and Turner 1992, Moran et al. this issue, Rudel 1989, Sanderson in press).

In this article, we propose an interdisciplinary approach for analyzing tropical deforestation in the Brazilian Amazon. We review both the physical and human dimensions of Amazonian deforestation and discuss some issues of measurement and analysis facing both physical and social sciences. We emphasize the need to analyze this problem across different scales of space and time, including local dynamics at the level of individual farms, regional patterns, and international conditions that influence Amazonian deforestation.

In the first part of this article, we take an empirical view, showing how deforestation can be measured and how satellite remote sensing can play an important role. Sociodemographic and economic data from standard census sources can supplement remote-sensing data to provide additional information. From this data-intensive set of observations and measurements, we propose an explanatory model in the second part. Here, we consider the relationship among deforestation and large-scale social, economic, and institutional factors. This discussion, leading from measurements to analysis, forms the basis for a research design presented in the last section. We center our discussion on the period from the late 1970s to the late 1980s, a period of most rapid change.

Physical dimensions of deforestation in the Amazon

To measure deforestation in the Brazilian Amazon, satellite remote sensing provides the best source of information. Landsat, Spot, and other sensors can be used to develop detailed maps of the rate and geographical extent of deforestation in tropical forests, and thus document the location and expansion of deforestation over time. It is also possible to use satellite data from the NOAA series of weather satellites to locate areas of intense deforestation. However, data from these sensors are too coarse to quantify precisely the areas or rates (Skole et al. in press).

Regional-scale patterns using remote sensing. We have mapped the area of deforestation in 1988 and the rate of deforestation between 1978 and 1988 for the Brazilian Amazon (Skole and Tucker 1993). We began with 210 Landsat Thematic Mapper images for the entire Legal Amazon of Brazil for 1988. Individual scenes were digitized using visual interpretation and standard vector geographic information system (GIS) techniques. The exact boundary between intact forest and deforested land was digitized in the universal transverse mercator projection at 1:500,000 scale. All areas of closed canopy forest that had been deforested by 1988 were delineated, including areas of secondary growth on abandoned fields and pastures when visible. Individual digitized scenes were projected into geographic coordinates (latitude, longitude), edge-matched, and merged in the computer to form a single, seamless dataset for the entire Legal Amazon. This dataset was then projected into a sinusoid equal-area projection to create the final digital map from which all calculations of area were made.

This analysis provides rates of deforestation lower than previously estimated. By comparison to 1978 (Tardin et al. 1980), we estimate the rate of deforestation to be 15 x 10(3) to 20 x 10(3) km/yr, which is considerably lower than estimates made without remote sensing data (Myers 1991) or from trend extrapolations (Fearnside 1982). Our estimates are in close agreement with those reported by the Brazilian Instituto de Pesquisas Espaciais (INPE 1992).

A digital map (page 317) was derived from satellite data of the average annual rate of deforestation in the Legal Amazon between 1978 and 1988. For simplicity of display at a small map scale, the data have been aggregated into 16 x 16-kilometer grid cells. Spatially explicit maps such as these can provide insights into large scale geographic patterns and trends in deforestation in the Amazon basin of Brazil on decadal time scales. Most deforested land is concentrated in a crescent along the southern and eastern fringe of the Amazon. We estimate the total area deforested in 1988 to be 230,324 km(2). This estimate suggests that 6% of the closed forests have been cleared to date over the entire Legal Amazon. This fraction is somewhat higher in certain states. In Maranhao, for instance, as much as 27% of the forest cover has been converted. Forest clearing in this part of eastern Amazonia has been occurring since the earliest settlements in the last century. However, 11.5% of the forests of Rondonia have been cleared, where there was little deforestation before the mid 1970s.

Figure 1. Land cover conversion in Brazil in (a) 1970, (b) 1975, and (c) 1980 was estimated using land census data. When incorporated spatially in a geographic information system, these data show the spread of agriculture, and hence deforestation, into the Amazon. Also shown are the high-density areas of large-scale agriculture in the south of Brazil. The map shows the density of natural land cover converted to agriculutre, expressed as a percentage (%) of the area.

Regional-scale patterns from land census data. It is possible to obtain deforestation estimates using tabular summaries from standard government census sources. In the example below, data for each of the 3973 town-level political districts in Brazil are obtained from the Brazilian Census of Agriculture ( IBGE 1970b, 1980). These data do not directly report the area deforested, but instead they provide estimates of land in various forms of permanent and temporary agriculture, including pastures. Because the major cause of deforestation in the Amazon is agricultural expansion (Moran et al. page 329 this issue), these data can be used as a proxy for deforestation, but they cannot provide a direct measure of deforestation as with remote sensing. One value of agricultural census reports is that they provide data not available from remote sensing, such as crop type, farm size, fertilizer use, and other information related to land use, management, and tenure.

Land census data can be mapped using a geographic information system. A digitized map of the political borders of each district (municipio) in Brazil was constructed from base maps ranging in scale from 1:1,000,000 to 1:2,000,000. Each polygon represents a political district and is the basic geographic unit for which the land-use data were tabulated. The area of various land-use categories was assigned to each polygon, resulting in maps showing the geographic distribution of landcover conversion (Figure 1). Because all data has been digitally encoded, they provide a spatial and temporal database for quantitative analysis of regional patterns. These maps show at three different times the areas that had been converted to agriculture expressed as fractions of the total area. The geographic pattern is similar to that derived from satellite data (map page 317).

Before 1970, most of the conversion had occurred in southern Brazil and along its coast. Different areas experienced conversion after 1970. By comparing maps, one can see that most of the changes in land cover between 1970 and 1980 occurred along two major fronts. One is a north-south corridor along the Belem-to-Brasilia Highway. The other area extends west from the state of Mato Grosso into the colonization areas in the state of Rondonia.

Inspection of the GIS-based dataset indicates that significant movement into the Legal Amazon began between 1975 and 1980. In 1970, only 14.6% of all agricultural land in Brazil (including pasture) was in the Legal Amazon. By 1980, this fraction increased to 22.3%. Between 1970 and 1975, 45% of the new conversions occurred in the Legal Amazon. Between 1975 and 1980, 56% of the new conversion occurred within the Legal Amazon, reflecting the beginnings of the trend toward expansion of agriculture and economic development in the region. A comparison of census-derived and satellite-derived estimates of the total area deforested is shown in Figure 2. The estimates are comparable, and they suggest an important role for census data to extend the historical record to periods before remote-sensing data were available.

Local-scale dynamics. Region-wide patterns of deforestation are the result of many local activities. Net deforestation is the sum of several gross land cover transitions: primary forest conversion, abandonment of agricultural land (which imitates secondary succession), and re-clearing of successional vegetation. These fine spatial and temporal scale dynamics are important because the pattern and timing of clearing and abandonment affect biogeochemistry and other physical processes.

We have described typical local-scale patterns of deforestation in the state of Rondonia, where many of the new Amazon settlements have been developed. The patterns are derived from an analysis of high-resolution satellite data (Spot multispectral data with pixel size of 400 m(2)) in a 642 x 10(3) ha study area for the years 1986, 1988, and 1989 (Skole 1992).

Figure 2. The total area deforested in the Legal Amazon, closed forest region, estimated from agricultural census records and satellite data combine to create an historical time series. Values are 10(6) ha.

We classified the data into three land-cover classes: intact forest, active agriculture from new deforestation, and second-growth vegetation following abandonment; these three classes of land cover are distinguished by the relative reflectance in the near-infrared and visible bands (Figure 3). Moran et al. (page 329 this issue) describe a similar pattern spectral discrimination. Extensive field verification of the classification took place in 1988, 1989, 1991, and 1993. This analysis thus provides a measure of the separate transitions between land-cover classes. Most analyses reported to date consider only forest and deforested areas, with the deforested area as the combined area of active agriculture and secondary growth. In addition, because the satellite data can be spatially registered, we have tabulated transition sequences for each 400-square meter parcel of land or individual field.

Figure 4 shows land-cover transitions that occurred in the study area between 1986 and 1988 and between 1988 and 1989. The values are annual transition rates for each period. Between 1986 and 1988, new agricultural land came from clearing 4.12 x 10(3) ha/yr of primary forest and 1.97 x 10(3) ha/yr of secondary vegetation. Between 1988 and 1989, 8.63 x 10(3) ha/yr of primary forest and 6.21 x 10(3) ha/yr of secondary vegetation were cleared for agriculture. Clearing of secondary vegetation is an important source of new agricultural land; between 1988 and 1989, 42% of the new agricultural land was created from clearing of secondary growth. The amount of agricultural land that remained as active agriculture from one year to the next was 10.06 x 10(3) ha/yr between 1986 and 1988 and 25.04 x 10(3) ha/yr between 1988 and 1989.

The area in secondary succession is significant. Of the total land deforested (active agriculture plus secondary growth), approximately 33% was secondary vegetation in 1989, an increase over the 25% measured in 1986. The large and increasing proportion was also shown in sites in the eastern Amazon by Moran et al. (page 329 this issue). These satellite observations agree with long-standing field observations, but heretofore there have been no quantitative measurements. Our ongoing analysis of the entire Legal Amazon using satellite data suggests that secondary vegetation is widespread throughout the region. The implications for carbon storage are important, because regrowing vegetation accumulates carbon previously lost to the atmosphere from clearing.

The turnover of secondary vegetation (i.e., abandonment and re-clearing) is an important process. In this study area, the area abandoned each year was 70% of the primary forest area cleared between 1986 and 1988, and 83% between 1988 and 1989. Approximately 11% of the active agriculture area was abandoned each year between 1986 and 1988. However, between 1988 and 1989, when there was more than a twofold increase in forest clearing, 22% of the agricultural land was abandoned annually.

Deforestation in the closed forest zone of the Brazilian Amazon Basin in 1988 was estimated from Landsat Thematic Mapper data. The map shows the total area deforested as of 1988. The orginal data were analyzed and digitized in a geographic information system at 1:250,000 scale as shown in the inset. For display purposes the data have been summed into grid cells of 16 km by 16 km, and represented as a density (precentage of the cell deforested).

If one-fifth of the agricultural land is abandoned each year, we estimate an average steady-state turnover time of approximately five years. This figure is generally consistent with what other observers have reported (Buschbacher 1986, Buschbacher et al. 1988, Uhl et al. 1988); land fertility and productivity decline to the point that the farmer abandons the land after approximately five years. Because satellite observations make it possible to separately track each 400-square-meter piece of land, it is possible to determine that out of the 5.804 x 10(3) ha that were abandoned to secondary growth between 1986 and 1988, 45% was re-cleared during the next year.

This analysis suggests the important and apparently inseparable coupling between land in active agriculture and secondary growth. The mode of production in this area is predicated upon maintaining both classes of land use. Moreover, abandonment rates tend to increase when increases in primary forest clearing produce net increases in secondary-succession area. Thus, local ecological conditions, methods of agro-ecosystem resource management, and local decision making are as much a driving factor in deforestation as are demographic factors.

Human dimensions of deforestation in the Amazon

Satellite observations provide an objective and quantitative approach to the measurement of deforestation. When mapped over large regions such as the Amazon Basin, it is possible to see geographical patterns. By analyzing the geometry, size, and spatial arrangement of clearings, satellite imagery can also be used to categorize different types of deforestation (e.g. large pastures, small farms, extraction activity, and mining). But these observations alone cannot completely identify why deforestation occurs or determine what factors influence regional trends or local dynamics. Economic and institutional factors caused the explosive rates of deforestation in Amazonia beginning in the middle of the 1970s.

Demographic Factors. In recent years, there has been a focus on the relationship between population and land use change. For example, Allen and Barnes (1985) surveyed population and deforestation data for 76 tropical countries using statistical correlation. They also examined multiple regressions of deforestation against other variables such as arable land, roundwood production, and gross domestic product. Their analysis suggested a low, but significant, correlation between population growth rates in the period 1970 to 1978 and deforestation reported for the period 1975 to 1980 from the FAO Forest Assessment (Lanly 1982). They concluded that population growth was the cause of deforestation globally.

In the Brazilian Amazon, Reis and Margulis (1990) have examined several anthropogenic causes for large and increasing deforestation rates during the mid-1980s, ultimately relating these rates to emissions of carbon dioxide. One conclusion they draw is that population growth relates to deforestation when population density is plotted against deforestation density (fraction of an administrative district deforested). A multiple regression model, relating deforestation to several anthropogenic variables including the density of population, yielded a high correlation coefficient (r(2) = 0.8). Kummer (1992) notes, however, that Reis and Margulis assume in their analysis that areas in question were 100% under forest cover at the beginning of their study period. This assumption, which is not always correct, has the effect of attributing all deforestation to a single time period.

Figure 3. A subset of the satellite data used to analyze deforestation and secondary growth turnover dynamics is displayed. Darker red areas are intact forest, blue areas are active agriculture (a), and bright red areas show secondary growth (b). The image is approximately 10 km by 10 km. The scene center was placed at S 9 31 03 (degrees, minutes, seconds) and longitude W 63 25 28 (degrees, minutes, seconds) and was acquired on 7/10/88

Regression analysis on deforestation rates derived from satellite data and population density for each of the several hundred municipal units in the Amazon for the period 1975 to 1978 revealed virtually no correlation at this scale (Skole 1994). The relationship reported by Allen and Barnes (1985) at the global scale is questionable when examined in smaller units.

Simple relationships to population growth may not alone describe factors driving deforestation in the Amazon. In a study of cattle ranching in the Amazon, Hecht (Hecht 1985, Hecht 1993), concluded that government policy, fiscal incentives, and the nature of individual farmer decisions in an inflationary economy (i.e., owning cattle is a good hedge against certain economic conditions) are more significant determinants of deforestation than are demographic considerations. A similar view has come out of recent studies of declining wood stocks in sub-Saharan Africa (Anderson 1986, Anderson and Fishwick 1984) and Southeast Asia (Kummer 1992). In these case studies, population growth is viewed as one variable in a multiple feedback system rather than a forcing function.

Economic and Institutional Factors. Deforestation in the Brazilian Amazon, particularly in the eastern part, has been occurring for a long time. However, our estimates of total deforested area derived from satellite and statistical records indicate that 90% of the deforested area in 1988 was created after 1970. We hypothesize that events and conditions at national and international scales influenced the explosive rates of deforestation during this initial period of Amazonian frontier development. Changes in the structure of the national and international economy and the emphasis of national development policies during this period have been the dominant influence on Amazonian deforestation.

The most important agent of deforestation in the Amazon was agricultural expansion, particularly for pasture. If one examines the geographical pattern of agricultural expansion in Brazil during the 1970s some interesting patterns emerge. Although the south of Brazil continued to have the highest density of agriculture in the late 1970s, its areal extent changed little. Agriculture quickly expanded into the Brazilian Amazon, where some states experienced the most dramatic changes anywhere in the country.

The state of Rondonia had nearly exponential deforestation rates during this period, as new colonization and settlement programs opened large tracts of forest. These settlement programs, and specific fiscal incentives were established in the 1970s as a way to encourage migration from overpopulated, poverty-stricken, and drought-ridden regions in the south and northeast of the country (Hecht and Cockburn 1989, Mahar 1989, Moran 1981, Schmink and Wood 1984). The vast Amazonia was seen by many as an empty frontier, which at once could be consolidated under Brazilian national sovereignty and provide opportunity for millions of poor and landless people (Bunker 1984a,b). Long after the establishment of settlement programs migrants continued to flow into the region (Lisansky 1990).

Figure 4. Transition rates between three land cover classes for (a) the period 1986 - 1988, and (b) the period 1988 - 1989. Values are hectares per year.

Deforestation in Amazonia is linked to changing demographic and economic conditions in the south of Brazil, particularly in the state of Parana. In the early 1970s, changes in land tenure and land use in Parana directly influenced deforestation in Rondonia. These changes resulted from international activities, particularly those related to world oil production, distribution, and price. Strong external factors created the preconditions for Amazonian deforestation that remained for a decade.

After the Organization of Petroleum Exporting Countries (OPEC) increased the price of oil in the mid-1970s, large amounts of oil money (petrodollars) flooded international money markets (Pool and Stamos 1987). The price of oil went from $1.30 per barrel in 1970 to $10.72 per barrel in 1974 and to $28.67 by 1980. Energy-dependent countries paid OPEC prices, resulting in a large net transfer of wealth from industrial economies to OPEC members. This revenue was then deposited in US and European banks in a process called petrodollar recycling (Pool and Stamos, 1987). Because banks must pay interest to depositors, these institutions were eager to find borrowers.

Developing countries such as Brazil required foreign capital to fund economic development, modernization, and industrialization programs. They also needed US dollars to pay for oil, because oil is bought and traded in dollars. Brazil's strategy appears to have been twofold: reduce the amount of imported oil by developing domestic sources such as hydroelectric energy, and borrow from foreign lenders to fund domestic economic development programs. Agricultural modernization was one focus of this development financing, because agricultural exports could be used to obtain foreign exchange and service the debt (Mahar 1989).

Figure 5. Allocation of crop credits in Brazil in 1978, as a percentage of crop area (World Bank 1982).

Agricultural modernization has been an important national goal in Brazil (Bunker 1984b, Mahar 1989, Moran 1981, World Bank 1982). In the last 20 years, total farmland area increased more than 60% and the land in crops increased 176% (IBGE 1970a,b, 1975, 1980, 1989). Between 1970 and 1980, there was large-scale financial investment in agriculture. Crop credits increased almost fivefold (World Bank 1982). The use of machinery and other commercial inputs grew as Brazil has become a leading agricultural exporter of such commodities as soybeans and oranges. Such investment programs were successful at least in part; rising crop credits occurred with increasing crop output, and the net value of agricultural output increased 2.68-fold between 1970 and 1980 (IBGE 1970a,b, 1975, 1980, 1989). By 1977, export crops made up more than 50% of the total value of principal crops.

Figure 6. Area planted over time in major crops in the state of Parana between 1965 and 1985.

The agricultural modernization programs resulted in changes in land allocation and land tenure. Figure 5 shows the distribution of crop credits by crop type in 1978. Three general patterns emerge. First, almost half of the total crop area receiving credits was used for three export crops: wheat, soybeans, and coffee. Second, the largest fraction was in soybean production. Third, little of the cropland receiving credits was used for staple crops such as black bean and manioc.

Figure 7. Change in farm size distribution in the state of Parana between 1970 and 1985. Units are the precentage increase or decrease in farms of each size class.

Soybean production was a major success story. The soybean area harvested increased sixfold in the 1970s, ten times more than any other crop except oranges and wheat. Soybean yields increased fivefold. The combination of land, fertilizers, improved seeds, and government-sponsored fiscal credits and incentives produced an internationally competitive export program.

Soybeans became one of Brazil's major export crops. Brazil provided 10% of the international soybean market in 1970, but half the global market by the early 1980s. Brazil had become the United States' chief competitor. Most of the soybean production was concentrated in two states, Parana and Rio Grande do Sul. Figure 6 shows the trend in area planted in some important crops in Parana during this period. Soybean production (and wheat) replaced coffee as the major crop in the region. Government programs concentrated on replacing coffee fields with soybeans (World Bank 1982), because the international market for coffee, unlike soybeans, was highly variable and undependable.

One reason for Brazil's competitiveness in the soybean market might be related to comparative costs of production. Although costs of such inputs as fertilizers and pesticides are high in Brazil compared with other countries such as the United States, land costs are half that of the United States (World Bank 1982). As a result, overall costs are lower for Brazilian production than in the United States.

Figure 8. Change in the out-migration rate for each state in Brazil between 1970 and 1980 were important for Parana state. Units are perentage change between 1970 and 1980. The state with the highest change in rate of out-migration was Parana (20). Other states shown: Rondonia (1), Acre (2), Amazonas (3), Roriama (4), Para (5), Amapa (6), Maranhao (7), Piaui (8), Ceara (9), Rio Grande do Norte (10), Paraiba (11), Pernambuco (12), Alagoas (13), Sergipe (14), Bahia (15), Minas Gerais (16), Espirito Santos (17), Rio de Janeiro (18), Sao Paulo (19), Santa Catarnina (21), Rio Grande do Sul (22), Mato Grosso (23), Goias (24), Distrito Federal (25).

The modernization of agriculture came with structural changes in the economy. A labor-intensive, small-scale agricultural system was being transformed into an important energy- and machinery-intensive component of the national economy, particularly in the state of Parana. Land prices rose significantly (World Bank 1982) as land was consolidated into larger holdings. Coffee, a labor-intensive crop, was replaced by soybeans and wheat, whose cultivation uses machinery. This transformation of land use changed land tenure. Figure 7 shows the change in farm-size distribution in Parana. There was a loss of small farms and an increase in large, presumably commercial and mechanized, farms.

The period 1970-1980 saw increased migration from rural to urban areas. Part of this migration was in response to increased opportunities and wages in urban areas. At the same time, commercialization and mechanization of agriculture in Parana displaced laborers. Migration out of the state of Parana during this period was higher than that of any other state (Figure 8).

In addition to the many migrants to urban areas, a large number also went to the Rondonia frontier (Hecht and Cockburn 1989). On the frontier, new migrants cleared forests, planted crops, and opened new pastures, confronting the Amazonian ecosystems as best they could. At the level of the town or farm, local dynamics of land use and land management are determined by coping strategies. For instance, Hecht (1993) argues that clearing for pasture by farmers (the most important land use resulting from deforestation) is a rational individual economic strategy and an efficient way to capture value from the land.

In the last few years, many of the government programs that promoted deforestation have been drastically curtailed. As these programs were restructured and debt tightened the availability of money, the rate of deforestation has declined. The latest estimates of deforestation in the Amazon suggest that the rate is now half of what it was in the late 1980s (INPE 1992, Skole and Tucker 1993).

An interdisciplinary approach

The problem of land cover change is complex and cuts across many scales of analysis. In the case of deforestation in the Brazilian Amazon, regional trends are influenced by large-scale external forces but mediated by local-scale conditions. A three-level, interdisciplinary approach to the study of deforestation in Brazilian Amazonia could be taken.

The overall approach would start with direct measurements of the rate, location, spatial pattern, and temporal characteristics of deforestation. Satellite remote sensing is a promising tool for objectively making these measurements at different spatial and temporal scales, from large-scale assessments of regional trends to local-scale analysis of complex dynamics. Although regional-scale satellite data alone might form the basis for empirical models with limited predictive capability (e.g., spatial trend or diffusion modeling), mapping of deforestation at scales of 1:100,000 to 1:500,000 with remotely sensed data would establish the regional context for integration with sociodemographic data from agricultural and demographic census documents.Such integration would provide useful information on land use, tenure, and management.

At a second level of analysis, case studies and field investigations could be carried out in conjunction with multitemporal, high-resolution satellite data at 1:50,000 to 1:100,000 scale to gain insight into local-scale dynamics of deforestation, abandonment, and second growth turnover. These case-study analyses would use survey research and statistical data from census documents to define the parameters that control local land-use strategies, which would in turn illustrate how changes in land use affects changes in land cover. Complementary to this view is the work of Moran et al. (page 329 this issue) and of Hecht (1993). Both describe the important role local conditions play in determining land use and individual economic strategies.

Because the causes of deforestation may also significantly relate to external institutional and economic factors, an elucidation of driving forces cannot be made with satellite data and field studies alone. In Brazil, the factors responsible for deforestation in the Amazon originated far outside the region. They involved land-tenure changes in the south of Brazil and changes in the rapidly developing national economy, to some extent catalyzed by excess petrodollars and international lending. The substitution of machinery for labor, which was an important component of national agricultural modernization efforts, affected the forests of the Amazon. Migration to the Amazon was partly a response to conditions and processes far removed from Rondonia and not solely the product of Brazilian population growth. Thus, deforestation is a more complex problem than simply there being too many people.

Political, institutional, and economic forces establish and modulate long-term conditions. Thus, it would be necessary at a third level of analysis to define the large-scale external factors and conditions that influence deforestation in the Amazon. The causes of Amazonian could then be considered deforestation in an international context.

Acknowledgments

This work was supported by National Aeronautics and Space Administration's Mission to Planet Earth, the EOS Data and Information System's Landsat Pathfinder Program, and the Department of Energy's Carbon Cycle Program. We acknowledge the useful comments on the manuscript from Robert Harriss and three anonymous reviewers, and insights from members of the IGBP Land Use/Land Cover Core Project Planning Committee.

References cited

Allen, J. C., and D. F. Barnes. 1985. The causes of deforestation in developing countries. Annals of the Association of American Geographers 75: 163-184.

Anderson, D. 1986. Declining tree stocks in African countries. World Development 14: 853-863.

Anderson, D., and R. Fishwick. 1984. Fuelwood Consumption and Deforestation in African Countries. World Bank, Washington, D.C.

Browder, J. O. 1988. Public policy and deforestation in the Brazilian Amazon. In R. Repetto and M. Gillis, eds. Public Policies and the Misuse of Forest Resources. Cambridge University Press, New York.

Bunker, S. G. 1984a. Modes of extraction, unequal exchange, and the progressive underdevelopment of an extreme periphery: the Brazilian Amazon, 1600-1980. American Journal of Sociology 89: 1017-1064.

___________. 1984b. Underdeveloping the Amazon: Extraction, Unequal Exchange, and the Failure of the Modern State. University of Illinois Press, Champaign .

Buschbacher, R. 1986. Tropical deforestation and pasture development. BioScience 36: 22-28.

Buschbacher, R., C. Uhl, and A. S. Serrao. 1988. Abandoned pastures in eastern Amazonia.II. Nutrient stocks in the soil and vegetation. J. Ecol. 76: 682-699.

Crutzen, P.J., and M.O. Andreae. 1990. Biomass burning in the tropics: impacts on atmospheric chemistry and biogeochemical cycles. Science 250: 1669-1678.

Fearnside, P. M. 1982. Deforestation in the Brazilian Amazon: how fast is it occurring? Interciencia 7: 82-88.

Food and Agricultural Organization of the sources assessment. FAO Forestry Paper 112, FAO, Rome, Italy.

Hecht, S.B. 1985. Environment, development and politics: capital accumulation and the livestock sector in Eastern Amazonia. World Development 13: 663-664.

___________. 1993. The logic of livestock and deforestation in Amazonia. Bioscience 43: 687-695.

Hecht, S.B., and A. Cockburn 1989. The Fate of the Forest. Verso, London, UK.

Houghton, R.A. 1991. Tropical deforestation and atmospheric carbon dioxide. Climatic Change 19: 99-118.

Houghton, R.A., and D.L. Skole. 1990. Carbon. Pages 393 - 408 In B. L. Turner, ed. The Earth Transformed by Human Action. Cambridge: Cambridge U.P.

Instituto Brasileiro de Geografia e Estatistica (IBGE). 1970a. Anuario Estatistico do Brasil - 1970b. IBGE, Rio de Janeiro, Brazil.

___________. 1970b. Censo Agropecuario - VIII Recenseamento Geral do Brasil. IBGE, Rio de Janeiro, Brazil.

___________. 1975. Anuario Estatistico do Brasil - 1975. IBGE, Rio de Janeiro, Brazil.

___________. 1980. Censo Agropecuario - IX Recenseamento Geral do Brasil. IBGE, Rio de Janeiro, Brazil.

___________. 1989. Anuario Estatistico do Brasil - 1989. IBGE, Rio de Janeiro, Brazil.

Instituto Nacional de Pesquisas Espaciais (INPE). 1992. Deforestation in Brazilian Amazonia. INPE, Sao Jose dos Campos, Brazil.

Kummer, D. M. 1992. Deforestation in the Postwar Philippines. University of Chicago Press, Chicago, IL.

Lanly, J. P. 1982. Tropical Forest Resources. Food and Agricultural Organization of the United Nations, Rome.

Lisansky, J. 1990. Migrants to Amazonia: Spontaneous Colonization in the Brazilian Frontier. Westview, Boulder, Colo.

Mahar, D. J. 1989. Government Policies and Deforestation in Brazil's Amazon Region. World Bank, Washington, D.C.

Meyer, W. B., and B. L. Turner. 1992. Human population growth and global land-use/cover change. Annu. Rev. Ecol. Syst. 23: 39-61.

Moran, E. F. 1981. Developing the Amazon. Indiana University Press, Bloomington.

Moran, E. F., E. Brondizio, P. Mausel, and Y. Wu. 1994. Ingtegrating Amazonian vegetation, land-use, and satellite data. BioScience 44: 329-338.

Myers, N. 1991. Tropical forests: present status and future outlook. Climatic Change 19: 3-32.

Pool, J. C., and S. Stamos. 1987. The ABC's of International Finance : Understanding the Trade and Debt Crisis. Lexington Books, Lexington, MA.

Reis, E. J., and S. Margulis. 1990. Economic perspectives on deforestation in Brazilian Amazon. Project Link Conference, Manilla, The Phillipines.

Rudel, T .K. 1989. Population, development, and tropical deforestation: a cross national study. Rural Sociology 54: 327-338.

Salati, E., and P. B. Vose. 1984. Amazon basin: a system in equilibrium. Science 225: 129-138.

Sanderson, S.1994. Political-economic institutions. In W. B. Meyer and B. L. I. Turner, eds. Changes in Land Use and Land Cover: A global Perspective. Cambridge University Press, New York.

Schmink, M., and C. H. Wood. 1984. Frontier Expansion in Amazonia. University of Florida Press, Gainesville.

Shukla, J., C. Nobre, and P. Sellers. 1990. Amazon deforestation and climate change. Science 247: 1322-1325.

Skole, D.L. 1994. Data on global land cover change: acquisition, assessment, and analysis. In W. B. Meyer and B. L. Turner, ed. Changes in Land Use and Land Cover: A Global perspective. Cambridge University Press, New York.

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

Skole, D. L., W. A. Salas, and L. Andres. in press. The use of AVHRR data for land cover classification. In A. Belward and J.-P. Malingreau, eds. Advances in the Use of AVHRR Data for Land Applications, Kluwer Academic New York.

Tardin, A. T., D. C. L. Lee, R. J. R. Santos, O. R. Ossis, M. P. S. Barbosa, M. L. Moreira, M. T. Pereira, D. Silva, and C. P. Santos Filho. 1980. Subprojecto Desmatamento. IBDF/CNPQ-INPE, Instituto de Pesquisas Espaciais, Sao Jose dos Campos, Brasil.

Uhl, C., R. B. Buschbacher, and E.A.S. Serrao. 1988. Abandoned pastures in eastern Amazonia. I. Patterns of plant succession. J. Ecol. 76: 663-681.

World Bank. 1982. Brazil: A Review of Agricultural Policies; a World Bank Country Study. World Bank, Washington, D.C.