• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Land-Cover and Land-Use Change in the Brazilian Amazon: Smallholders, Ranchers, and Frontier Stratification

Posted on: Thursday, 7 September 2006, 09:20 CDT

By Aldrich, Stephen P; Walker, Robert T; Arima, Eugenio Y; Caldas, Marcellus M; Et al

Abstract:

Tropical deforestation is a significant driver of global environmental change, given its impacts on the carbon cycle and biodiversity. Loss of the Amazon forest, the focus of this article, is of particular concern because of the size and the rapid rate at which the forest is being converted to agricultural use. In this article, we identify what has been the most important driver of deforestation in a specific colonization frontier in the Brazilian Amazon. To this end, we consider (1) the land-use dynamics of smallholder households, (2) the formation of pasture by large-scale ranchers, and (3) structural processes of land aggregation by ranchers. Much has been written about relations between smallholders and ranchers in the Brazilian Amazon, particularly those involving conflict over land, and this article explicates the implications of such social processes for land cover. Toward this end, we draw on panel data (1996-2002) and satellite imagery (1986-1999) to show the deforestation that is attributable to small- and largeholders, and the deforestation that is attributable to aggregations of property arising from a process that we refer to as frontier stratification. Evidently, most of the recent deforestation in the study area has resulted from the household processes of smallholders, not from conversions to pasture pursuant to the appropriations of smallholders' property by well-capitalized ranchers or speculators.

Key words: land-use change, land-cover change, frontier speculation, Brazilian Amazon, political ecology, farming systems, property dynamics, ranching.

Deforestation in the Amazon basin is a process of land-cover and land-use change that has attracted a great deal of interest and generated controversy. Early explanations held that corporate interests were responsible for most forest loss in the region and that poor farmers who migrated there in response to the government's colonization programs played a marginal role because of the scale of operation, primitive technology, and focus on food production that characterizes many smallholders (Wood 1983; Hecht 1985, 1988). Recently, a more nuanced perspective has emerged, based, in part, on field observations that small-scale agriculture in the Amazon can be economically viable (Moran 1984; Hecht 1993; Faminow 1998; Walker, Perz, Caldas, and Silva 2002; Walker 2003). According to this perspective, in addition to well-capitalized ranchers, sizable populations of poor colonists now survive in Amazonia, and their farming practices account for an appreciable amount of forest loss. Despite growing interest in colonists' deforestation processes, their importance in relation to those of largeholders remains unresolved.

This article addresses deforestation that is associated with smallholders and largerscale commercial interests, typically ranchers. It goes one step further by considering, in addition to their land-cover manipulations in isolation, the impacts stemming from their social relations involving the possession and dispossession of landed property. The article identifies what has been the most important driver of deforestation in a specific colonization frontier in the Amazon basin: (1) the land-use practices of small-holders, (2) pasture formation by large-scale commercial interests, or (3) the turnover of property, specifically aggregations (or consolidations) of smallholdings that bring about accelerated deforestation.

Pursuant to this objective, we present analyses that are based on information from a cadastral survey of a panel of smallholders, collected in 1996 mid 2002, which enabled us to track property- ownership dynamics, and analyses of satellite images of land-cover change from 1986 to 1999. The study area was the municipio (county) of Uruar, located on the Transamazon Highway in the State of Par, Brazil (see Figure 1). Uruar was opened to settlement in the early 1970s with road construction and federal government policies that were aimed at colonization (Smith 1982; Moran 1984). Since the construction of the Transamazon Highway, smallholders and ranchers, as well as loggers, have been actively deforesting the region (Bunker 1981; Smith 1982; Moran 1984; Walker 2003).1 While other Amazonian frontiers have experienced a greater impact from large- scale ranching, the Uruara study site is in a region where federally promoted, smallholder colonization has been pronounced (Walker, Moran, and Anselin 2000).

Much has been written about social relations, especially contentious relations, between smallholders and ranchers in the Brazilian Amazon (Schmink 1982; Schmink and Wood 1992; Faminow 1998; Porro 2002). This article focuses on these groups' competition for land and empirically analyzes both the intensity of the aggregation of property in the study area and its impacts on forest cover. Thus, we frame Amazonian deforestation in the context of heterogeneous agents who have produced a structure of land distribution that may generate environmental change. Our analysis considers the dynamics of land cover, land use, and property ownership in a single area and mostly neglects the broad-scale forces that have emerged from the macroeconomy and from the distribution of wealth at the national scale. Nevertheless, development in the Amazon has clearly been conditioned by such forces, and we do not neglect them in the final discussion of our empirical findings. We recognize that loggers, miners, and government bureaucrats have also played important roles in the loss of the Amazon forest (Hecht and Cockburn 1990; Smith, Serro, Alvim, and Falesi 1995). Thus, we deliberately restrict our focus to only one aspect of the regional problem.

Our presentation proceeds as follows. First, we discuss Amazonian deforestation and the role played by small- and largeholders, or ranchers. We also present our concept of frontier stratification, which we use to link the changes in the distribution of land, including the aggregation of property, to landcover and land-use change. Then we present empirical analyses of (1) deforestation occurring on small- and largeholder properties and of (2) property aggregation (and disaggregation) and resulting land-cover and land- use changes, followed by a discussion of the data in light of the issues under scrutiny. Finally, we place our results in the context of the greater effort to understand environmental change in Amazonia.

Smallholders, Ranchers, and Frontier Stratification

The narrative of Amazonian deforestation is well known and conforms generally to processes that have been observed in other parts of the tropics that are experiencing deforestation (Angelsen and Kaimowitz 1999; Geist and Lambin 2002; Gutman et al. 2004; Turner, Geoghegan, and Foster 2004). A federal government intent on securing national borders, staking claims to Tiatural resources, and seeking to slake the populist demands of displaced agricultural workers viewed the region as a solution to social, geopolitical, and economic problems (e.g., Simmons 2002). Consequently, investment in infrastructure was initiated in the late 1960s and throughout the 1970s with the construction of an extensive highway system crisscrossing the region (Smith 1982; see Figure 2). To complement these investments, the federal government also initiated colonization and incentive programs that attracted both people and capital to the Amazon region (Sawyer 1984; Schmink and Wood 1984; Browder 1988; Binswanger 1991; Schmink and Wood 1992; Alston, Libecap, and Schneider 1993; Porro 2002). Subsequently, market factors came into play, and deforestation in the Amazon basin developed momentum (Pfaff 1999; Anderson et al. 2002). The environmental consequences are apparent today, with a forest loss of about 17 percent in the Brazilian portion of the basin. This process has remained persistent over the past few decades and may even be accelerating (INPE 2005).

This broad overview leaves open more specific accounts of structure, agency, and environmental change. Appealing to notions of agency and structure, although consonant with political ecological theory, presents an interpretive challenge in the Amazonian case, however, where much of the discourse has relied on alternative languages. Consequently, we now set out our conceptual vocabulary. Toward this end, we first consider agents and structures and then address how the agents in question affect land cover in the study area. Next, we develop the concept of frontier stratification, an outcome of agents' interactions leading to the distribution of land resources, which drives a hypothesized deforestation process resulting from the aggregation of property. This conceptual framework sets the stage for the empirical analysis that follows.

As we indicated earlier, the agents of interest in this setting are land managers (Turner, Geoghegan, and Foster 2004), specifically small- and largeholders who manipulate types of land cover by virtue of their land-use decisions. Such agents function within the context of structures, which constrain-or enable-the actions they take; they also contribute to the formation of structures, particu\larly at local scales. In Amazonia, federal development policy and the national distribution of wealth have often been cited as drivers of deforestation; they clearly also represent macroscale structures. Central to our analysis are local structures of land distribution and the power relations that emerge endogenously by virtue of exchanges of property, including both aggregation and its obverse, disaggregation.2 The emergence of local land-distribution structures in the Amazon is often violent, and many deaths have resulted from related conflicts in the state of Par over the past few decades (Simmons, Perz, Pedlowski, and Silva 2002; Simmons forthcoming). The study area, Uruar, has remained relatively peaceful, however, and properties are exchanged through markets, not by violent acts or threats thereof.

Agents of Deforestation

Many agents have actively participated in the loss of the Amazonian forest (Binswanger 1991; Schmink and Wood 1992; Roberts and Thanos 2003; Walker 2003; Margulis 2004), but this article focuses on colonists and largeholders, both of whom clear forest to pursue agricultural activities. A considerable amount of research has addressed these groups' impact on land cover. Largeholders were frequently implicated in the widespread use and abuse of federal incentive programs during the early years of Amazonian development (Browder 1988; Yokomizo 1989; Hecht and Cockburn 1990; Binswanger 1991; Smith, Serro, Alvim, and Falesi 1995; Alston, Libecap, and Mueller 2000). These incentives attracted a substantial amount of corporate capital and led to much early deforestation; subsidized projects averaged 22,300 hectares in Mato Grosso, 12,400 in Gois, and 9,500 in Par (Yokomizo 1989).

Many early commentators argued that large-scale ranching was not economically rational in Amazonia and that large holdings existed only because of the incentives and land speculation (Mahar 1979, 1989a, 1989b; Hecht 1984, 1985; Browder 1988). Since then, the economic environment has changed with lower transportation costs to the south of the country and stable and growing prices for beef, given globalizing Brazilian markets (Arima and Uhl 1997). The creation of large ranches in Amazonia remains a desirable economic choice for those with capital (Arima, Walker, Perz, and Caldas 2005), although soy is making considerable inroads along the margins of the basin in drier cerrado (scrub-savannah) environments (Brown, Jepson, and Price 2004; Jepson 2005), where most current soy in the Amazon is produced (Nehmi Filho and Pusch 2003). Although some varieties of soy may eventually enable increased production in the wetter environments of forest areas, soy is not currently a major driver of landcover change in most of the forested areas of the basin. Nevertheless, the land-cover impacts of soy are fairly similar to those of pasture; both pastures and soy fields require the complete clearance of forest over extensive areas.

In the state of Par, more than 51 percent of agricultural lands are in holdings that exceed 1,000 hectares, and nearly half that land is controlled by enterprises that are larger than 10,000 hectares, many of them ranches (Simmons 2004). Be this as it may, the 1.5 million colonist households in the basin represent a demographic and ecological force to be reckoned with, given their number and agricultural activities, and much recent research has addressed their land-use practices. So-called life-cycle effects have received considerable attention in this regard. As described by Walker and Homma (1996), McCracken and colleagues (1999), Perz (2001), and Walker (2003), young smallholders with nonworking children typically minimize their risks by selecting food crops with annual yields, which entails a system of shifting cultivation and relatively minimal forest clearance. As children enter the household workforce, the family opts for commercial production with investments in perennials, ranching, or some combination. The choice of ranching, which is common, leads to extensive deforestation (Walker, Moran, and Anselin 2000).

Empirical research has suggested the co-evolutionary processes of smallholder households and agricultural strategies and the importance of traditional market factors in affecting smallholders' land-cover and land-use choices. Demographic attributes, including the size of the family (Rudel and Horowitz 1993; Pichn 1997), the number of male workers (Sydenstricker and Vosti 1993; Walker, Perz, Caldas, and Silva 2002), and the dependency ratio (Walker, Moran, and Anselin 2000) affect both the amount of deforestation and the choice of crops. Other household characteristics, such as wealth (Alston, Libecap, and Schneider 1993; Pichn 1997; Walker, Perz, Caldas, and Silva 2002), length of residence (Godoy, Wilkie, and Franks 1997; Pichn 1997; Godoy et al. 1998; Walker, Perz, Caldas, and Silva 2002), and access to markets (Ozorio de Almeida and Campari 1995; Godoy, Wilkie, and Franks 1997; Pichn 1997) have also been found to be significant variables. Evidently, demographic structure, wealth status, and market factors play important roles in the deforestation processes of smallholders (Reis and Guzmn 1994; Brondzio 1997; Pfaff 1999; Anderson et al. 2002; Walker 2003).

Smallholders and Ranchers: Relative Impacts

Agricultural activities and ranching clearly require forest clearance, but the relative contribution to deforestation in the Amazon basin that is accounted for by smallholders and large-scale ranchers remains an empirical question, one that this article addresses for the study area. In a study of the Brazilian portion of the Amazon basin, Fearnside (1993) estimated that large-scale ranching was responsible for 70 percent of all deforestation in 1990 and 1991. This proportion appears to be the upper limit of the estimates of the largeholder share, which ranges from as low as 25 percent (Faminow 1998). Walker, Moran, and Anselin (2000) showed that the relative degree of accountability is spatially variable. In a microregion in the south of Par, targeted by the Superintendncia do Desenvolvimento da Amaznia (that is, the Superintendency for the Development of the Amazon, or SUDAM), large enterprises accounted for 100 percent of the land cleared (in 1986), while along the Transamazon Highway, in the vicinity of the study area (see Figure 1), the proportion was 8 percent (in 1992).3 Yokomizo (1989) also found considerable variation at the state scale (in 1988), specifically for SUDAM-sponsored projects. In Mato Grosso, the proportion of deforestation that was associated with such enterprises ranged up to 21 percent, but during the same period in Par, it was only 7.5 percent (Yokomizo 1989).

Frontier Formation, Agent Interactions, and Deforestation

Although research has addressed the individual deforestation processes of smallholders and ranchers in isolation and these agents' relative contributions to deforestation, little is known about land-cover changes that result from landholders' interactions. Yet, social processes involving relations between agents have been identified as key to understanding certain types of deforestation, such as when colonists follow logging roads, causing further deforestation (Schmithusen 1980; Fearnside 1986; Schmink and Wood 1992; Myers 1993; Geist and Lambin 2002; Nepstad et al. 2002); Walker (2003) called specific attention to similar "process linkages" between smallholders and ranchers for the Amazonian case.4 Despite the lack of research in this regard, a substantial literature of high relevance to the issue is organized under various concepts of the frontier, which have played a significant role in describing the emergence of social structure and stratification in newly settled areas of Latin America and Brazil. These concepts consider the social welfare implications of the phasing of frontier occupation, typically with subsistence fanners in the vanguard and commercial interests following in their wake.

Perhaps preeminent among these concepts is the hollow frontier, first used to describe the dynamics of coffee production in the southern and central parts of Brazil (James 1969). In these areas, relatively wealthy largeholders displaced small-scale farmers through the aggregation of property and the largely forced exodus of smallholders, leading to rural depopulation and a demographically "hollow" frontier (James 1969; Taylor 1973; Casetti and Gauthier 1977).5 More recently, Foweraker (1981), Wood (1983), Messina and Walsh (2000), and Rudel (2002) described similar processes of property aggregation (cf. consolidation) and the displacement of agents in the Amazon region. As in the coffee frontier, well- capitalized interests, including land speculators and ranchers, consolidate the properties of subsistence farmers through market transactions or outright expulsions. The smallholders then move to unoccupied areas, while the land consolidators convert the subsistence farmlands into pasture (Casetti and Gauthier 1977; Wood 1983; Brown, Sierra, and Digiacinto 1994; Kudel 2002). Smallholders may also convert their holdings to pasture ahead of consolidating ranchers, in anticipation of increased land value and eventual sale (Foweraker 1981; Faminow 1998).

A second closely related concept is the speculative frontier, which refers to the occupation of land in the Amazon by capital interests who are intent on speculative holdings, as opposed to productive use (Foweraker 1981; Sawyer 1984; Hecht and Cockhum 1990; Schmink and Wood 1992; Alston 1996). Speculation has been a recurrent theme in the narrative of the economic development of the Brazilian Amazon, in which the federal government played a decisive role through the provision of incentive programs that attracted a great deal of capital to the region. As we discussed earlier, capital investments drawn by these incentive programs created huge holdings, and smallholders w\ho had been living on such newly acquired lands in the hopes of eventual possession were routinely evicted (Sawyer 1984; Schmink and Wood 1992).

Evidently, hollow and speculative frontiers entail a social interaction between smallholders and well-capitalized interests, the latter of whom possess either productive or speculative interest in the lands of the former, which they aggregate into large holdings. Thus, both frontier concepts provide accounts of an emergent social structure that arises by virtue of the interactions of heterogeneous agents in a process that we refer to as frontier stratification. Although the smallholders who initiate frontier settlement are poor, they share their poverty in relative equality until the aggregation of property causes the distribution of land to be skewed and drives social stratification (Moran 1989; Walker, Moran, and Anselin 2000; Coomes and Burt 2001; Walker 2003).

Even though theorists of frontier stratification have commented on its possible environmental implications (e.g., Wood 1983; Sawyer 1984; Pfaff 1999), interest in smallholder-largeholder interaction has focused mainly on the equity issues that are associated with dispossession of the poor (e.g., Foweraker 1981; Schmink and Wood 1992), a shortcoming that this article seeks to redress.6 The interactions as described (i.e., aggregation and dispossession) and their emergent social structure would seem to imply a land-cover outcome, given the different land-use strategies of smallholder colonists and ranchers. From a strictly behavioral perspective, if economic incentives that are based on scale economies drive ranchers to consolidate smallholdings to increase pasture area, deforestation intensifies (Walker, Moran, and Anselin 2000; Sierra, Tirado, and Falacios 2003). Despite considerable attention by political ecologies to the role of social structure in environmental degradation (Blaikie and Brookfield 1987; Hall 1989; Bryant 2001; Zimmerer and Bassett 2003; Bobbins 2004), research on deforestation in the Amazon basin has paid little empirical attention to impacts that may be associated with frontier processes leading to land consolidation and social stratification.

Land Cover, Land Use, and Property Ownership

Clearly, both large- and smallholders are found throughout the Amazon basin, and their interactions present both social and environmental concerns. This article addresses the environmental aspects of these interactions by considering the extent to which the aggregation (cf. consolidation) of property has triggered conversions to pasture in the study area. The interactions we consider are restricted to those involving land exchanges, namely, aggregations and disaggregations of properties, which are observable in our data. In addition, the article addresses the deforestation that is attributable to individual agents acting independently of one another. Thus, we seek to determine the importance of frontier stratification to the study area's land-cover change processes over a recent period. We do so by measuring deforestation that is associated with the aggregation of property and by comparing it to the in situ processes occurring on both small and large properties.

Data and Methods

The nature of the posed research questions requires an empirical analysis in two parts. First, we consider the extent of deforestation by small- and largeholders independently, using remotely sensed data derived from Landsat TM and ETM images, as shown in Figure 3. The town of Uruar is centered close to the corner of four Worldwide Reference System II (WRSII) scenes (Path 226, Row 62; Path 226, Row 63; Path 227, Row 62; Path 227, Row 63), which necessitated the creation of a mosaic, the seams of which are faintly observable upon close inspection of the figures. To obtain land-cover data for the analysis, we produced four classified mosaics for 1986, 1988, 1991, and 1999, as presented in the four panels of Figure 3. The Landsat TM (data from 1986, 1988, and 1991) and ETM (data from 1999) images were individually classified, and deforestation was measured for small- and largeholders individually over time.

We used a hybrid classification method and identified five land- cover categories: water, cloud, cloud shadow, forest, and not forest. The hybrid classification begins with an unsupervised algorithm that identifies 255 classes, which are then evaluated for spectral separability using "transformed divergence" to eliminate overlapping classes. Transformed divergence is a measure of spectral class dissimilarity. The calculation of transformed divergence requires more than one category. For example, if two classes are compared and their transformed divergence values are high, the two classes are highly dissimilar. Messina and Walsh (2000) used a similar hybrid classification using transformed divergence to assess class separability (dissimilarity). The goal of this stage of the process is to pare down the number of categories so that each is spectrally dissimilar. Once the unsupervised classification is evaluated and all overlapping categories are eliminated, the raw image is classified using the spectral characteristics of the remaining classes in a supervised routine. The resulting image is then spatially filtered to eliminate small areas, resulting in more contiguous land-cover units. This step of the classification is necessary to avoid checkerboard land-cover patterns resulting from surface anomalies, such as exposed soils within forest areas or rock formations. After small areas are removed, each aforementioned land- cover class is assigned to one of the land-cover categories. The individually classified images are then mosaicked to create a classified land-cover map for the entire study area for each year.

The second part of the analysis addresses potential interactions between small- and largeholders by examining the dynamics of the aggregation (consolidation), disaggregation, and fragmentation of property; household dynamics for individual properties are also considered, addressing both the farming system and demographic change. These two analyses are enabled by panel data for a sample of lots and properties in the study region covering the six-year period between 1996 and 2002. In our terminology, the lot is a physical unit, a 100-hectare piece of land, which was demarcated by the National Institute for Colonization and Agrarian Reform (INCRA) for smallholders in the early years of colonization. Lots are to be differentiated from glebas, which in the study area generally average either 400 or 3,000 hectares and were originally planned for occupation by well-capitalized agro-industrial interests (see Figure 4 for examples of glebas). By way of contrast, we define a property as an economic unit that is under the control of a single owner. As such, properties can consist of one or multiple lots, depending on how many the owner possesses. Figure 4 presents INCRA's cadastral plan from the early period of colonization, overlaid on the 1999 Landsat mosaic. The plan shows many small, elongated rectangles, organized along settlement roads heading north and south from the Transamazon Highway (BR-230). These rectangles are the 100-hectare lots (n = 3,263) originally marked by INCRA for smallholder colonization. A number of 3,000-hectare glebas (n = 9) are also evident as larger rectangles.

The panel data were collected in two phases. In the (North American) summer of 1996, researchers from the Brazilian agricultural extension agency (EMBRAPA), the University of Texas, and Florida State University visited 261 smallholder properties in Uruar municipio and implemented a questionnaire that gathered information on household characteristics and land uses. In the summer of 2002, another group of researchers from the Instituto do Homem e Meio Ambiente da Amaznia (IMAZON), the Federal University of Bahia, and Michigan State University returned to Uruar and attempted to visit the same smallholders who were interviewed in the 1996 survey to update the land-use and household information that was collected during the original survey effort. They managed to find 150 of the 367 original lots and 116 of the original farm properties.

Overall, the panel sample has a large random component within the cadastral plan, focusing on smallholders, not glebas. During preliminary fieldwork prior to 1996, initial survey teams attempted to sample on the basis of farmstead counts and distance, but these approaches proved intractable, given the owners' absenteeism and the difficulty of observing dwellings at specified kilometers. Ultimately, a principle of opportunity was observed, whereby as many interviews were obtained as possible in a daily period, on the basis of which households had members who were willing to be interviewed. Regular reference to a map ensured that the sample was spatially well distributed across the study area. Even so, a systematic bias may arise owning to the omission of households that are located far from the highway (BR-230); deep penetrations to include a larger number of remote properties in the sample proved problematic because of time constraints and difficulty of access. Although such households are typically the poorest and exist at the very margin of subsistence, they are not substantially different from those in the sample who live closer to the town of Uruar. Indeed, most small farmers in the area are close to subsistence, regardless of their proximity to BR-230.

Smallholder and Largeholder Deforestation Between 1986 and 1999

Deforestation by year and by size of holding is presented in Table 1. The size class breakdown is highly aggregate, drawing a distinction only between large glebas of 3,000 hectares (n = 9) and all other property sizes (n = 3,263), the vast majority of which are 100 hectares. The intermediate glebas, those with approximately 400 hectares, are included in the smal\lholder count, since some smallholder properties with multiple lots are the same size. The data show that, in general, the amount of deforested land in the two size classes increased over time. The degree of increment, however, is different, with large glebas adding only about 5 square kilometers over the 13-year period, from 33.7 to 38.6 square kilometers. It appears that deforestation associated with these highly capitalized interests occurred early in the colonization period during the 1970s7 and has been rather static since. By way of contrast, smallholder deforestation has steadily increased, more than doubling from 447.1 to 1,048.1 square kilometers over the same period.8 Furthermore, because the recent immigration and settlement of smallholders are not significant in the study area within the bounds of the cadastral map used for this analysis, large amounts of deforestation have not been caused by further settlement.

Land Distribution and Property Dynamics

The state of Par shows uneven patterns of land distribution, which worsened between 1985 and 1995. Gini coefficients for the state ranged from 0.71 to 0.82 over the period, a change reflected in the country at large, even with continuing development of the vast agricultural frontiers of the Amazon (Censo-Agropecurio 1995- 1996; Simmons 2004). Despite the pattern in the state at large, the distribution of landholdings in the Uruar municipio was relatively equitable in 1996, with a Gini coefficient of 0.18 (calculated from the CensoAgropecuirio 1995-1996).9

Panel data depicting the aggregation, disaggregation, and fragmentation of smallholders' properties in the region are presented in Table 2. In the present context, property aggregation occurred when single or, multilot properties that existed in the 1996 purvey were incorporated into larger, multilot properties in 2002. These multilot holdings can be very large, occasionally encompassing up to seven 100-hectare lots (Figure 5). Disaggregation occurred when multilot properties in the 1996 survey had fewer lots in 2002, while properties that did not change their number of lots were in "equilibrium." Fragmentation refers to the selling off of portions of individual lots. The data are broken down by ownership class (new or old owner), as well as by the magnitude of change.

The upper panel of the table summarizes 16 instances of aggregation. In 9 of these instances, single-lot properties from 1996 evolved into larger holdings, and in 7, aggregations incorporated multilot holdings from the early period. Old owners, those who held lands in the study area since before the 1996 data collection, appear to have been slightly more active than new ones, those who arrived in the study area after the 1996 survey. New owners account for 4 aggregations and old owners account for 12. Given that old owners are considerably more frequent in the sample, new owners appear to have been much more likely to aggregate. The actual magnitude of aggregations, as measured by the number of 100- hectare lots, is not particularly dramatic. Of the singlelot properties in 1996, five aggregated to two-lot properties, three to a three-lot property, and one to the six-lot holding with a new owner. Nor did aggregations of multilot properties from 1996 result in particularly large holdings, either. One four-lot property in 1996 was aggregated by a new owner into a seven-lot holding, the largest in the panel. In general, aggregations of multilot properties, relatively small to begin with, appear to have been incremental, only adding one or two lots.

The data show 16 instances of property disaggregation. In general, these are more dramatic than the sample aggregations in that for all hut one case, the result by 2002 was a single-lot holding. Although in 12 cases this disaggregation represented the sale of only one lot (e.g., from a two-lot to a onelot holding), the data show one disaggregation from six lots to one lot and one from five lots to one lot. During this period, disaggregations by old owners were much more common than those by new owners, of which there are 14 cases.

The dynamics at the lot level are much more common than those within lots. The panel reveals only eight cases of the fragmentation of lots, occurring more often among old owners than among new owners: four fragmentations were observed on singlelot properties, two on two-lot properties, and two on three-lot properties. The magnitude of lot fragmentation ranged from 5 to 50 hectares and in all cases left the owner with at least half of the original holding.

Household Dynamics

Farming Systems. Table 3 presents data on farm system dynamics, and Table 4 presents data on farm-yield change over the period in question. The information is organized by equilibrium, aggregating, and disaggregating subsamples. For each subsample, data are presented for the three main components of the smallholder systems, namely, annuals (com, rice, and beans), perennials (cocoa, black pepper, and coffee), and ranching (see da Veiga, Tourrand, and Quanz 1996). Other crops were grown, such as cassava and coconut, but greater effort was expended on the production of the annuals and perennials (Tables 3 and 4), and on ranching. Furthermore, fallow dynamics do not seem to have played a significant role in the aggregation or fragmentation of lots.

Two strong trends may be seen in the subsample of properties at equilibrium. First, areas planted and production magnitudes for annual crops diminished substantially during the period, with yields declining slightly for all annuals (see Table 4), accompanied by a decline in planted areas (see Table 3). The second notable trend was the increase of cattle herds. On average, equilibrium properties nearly doubled their herds, from 27.32 to 51.29 animals (see Table 3), and their stocking rates from 0.90 animals per hectare to 1.56 animals per hectare (see Table 4). That the amount of land in pasture increased only marginally (from 30 to 33 hectares) suggests that pasture was created in advance of animal stocking, the process of which is observable over a three-year period in Figure 5, possibly to secure land claims and as part of the long-term process of creating farms (Foweraker 1981; Faminow 1998; Walker 2003).

The dynamics of perennials production are mixed for properties at equilibrium and, on the whole, are not as strong as for annuals and ranching. The area dedicated to cocoa, black pepper, and coffee increased uniformly, but not by a large area. On the other hand, production increased for all perennial crops (see Table 3), although increases in cocoa production were not substantial. For properties at equilibrium, increases in coffee production were substantial, from 21.18 to 278.68 kilograms, partly because of greatly increased yields (from 43.22 to 188.3 kilograms per hectare, a 24percent increase). The decrease in the production of black pepper was at least its dramatic as the increase for coffee, falling from 959.3 to 327.01 kilograms, with parallel declines in yield, probably owing to continuing problems with fungal attack (fusarium spp.) and falling prices (Walker, Perz, Caldas, and Silva 2002). Generally, the trends in the production and area of annuals among equilibrium properties were similar for those undergoing aggregation and disaggregation dynamics, except for beans, which showed increments in production in all subsamples except aggregating properties (see Table 4). Com seems to have been an exception in disaggregating properties, since it also showed increments in production, possibly because disaggregating properties are more likely to focus on annuals, which are the primary food-pnxluction activity for struggling (i.e., disaggregating) properties. Changes in perennial activity are more difficult to generalize and had somewhat distinct patterns across the types of properties, although coffee production and area increased across all types of properties. Cocoa area decreased sharply for aggregating properties, from 7.32 to 3.35 hectares (see Table 3), a reduction that was somewhat attenuated for production (1,965 to 1,765 kilograms, owing to increased yields). On the other hand, the area of cocoa production grew for disaggregating properties, from 3.9 hectares to 6.49 hectares (see Table 3), while production fell (2,205 to 1,888.75 kilograms). The production of black pepper fell uniformly across the aggregating and disaggregating property types, and yields generally decreased, although there was actually an increase in area, albeit only from an average of .55 to .58 hectares.

Aside from coffee, ranching is the only activity that showed a consistent trend over the three property classes during the study period, with increments in pasture, herd size, and stocking rates observed for equilibrium, aggregating, and disaggregating properties (see Table 3). The magnitudes of change were somewhat variable, however, and most notable for aggregating properties, which had dramatic increments for both pasture area, from 42.75 to 114.8 hectares, and livestock, from 28.19 to 119.27 animals. Stocking densities also increased, despite the overall increases in pasture area.

Demographic Change. The demographic data in Table 5 reveal changes in family structure accompanying the evolution of the farm system over the study period. The average total population of properties decreased for all property types between 1996 and 2002, with the most marked decline observed for equilibrium properties. This decline is associated with declining population density, which was especially pronounced for aggregating properties (from .07 to .03 persons per hectare). Dependency ratios display a more variable pattern, increasing slightly on equilibrium properties (from .32 to .33 dependents per total property population), but decreasing on those that were aggregating (from .33 to .30) and disaggregating (from .36 to .35). Equilibrium and aggre\gating properties contained a growing number of elderly individuals, but fewer children, while disaggregating properties showed opposite trends for both. The use of day labor increased sharply across all three subpopulations.

Discussion

The analysis revealed the dynamics of land-cover change in an Amazonian region from 1996 to 2002 for two different agricultural agents (largeholders and smallholders). The image analysis showed that the largest holdings had not been aggressive in their alterations of vegetative cover since 1986. The glebas exceeding 3,000 hectares did not expand greatly since the early years of analysis, as can be seen by the glebas in Figure 4, or encroach into nearby smallholder properties. Over the 13-year period, the cleared lands on these properties increased by approximately 14 percent (from 33.77 to 38.63 square kilometers), or slightly more than 1 percent per year, on average. The data for smallholders provide a marked contrast, since smallholders more than doubled their cleared areas, from 447.13 to 1,048.08 square kilometers, or a 134percent increment. This is an average annual rate over the entire period exceeding 10 percent per year, 10 times greater than that observed on the glebas, which implies that the importance of smallholder fanning to the region's deforestation has been growing over time. Although these properties (including the midsized holdings of up to 400 hectares) accounted for 93 percent of the cleared lands in 1986, their share of total deforestation had risen to 97 percent by 1999.

The satellite image analysis provides insights into the role of heterogeneous agents in accounting for aggregate land-cover changes, but it cannot reveal the nature of their social interactions or the impact of these interactions on land cover. To do so, it is necessary to know if largeholders appropriated the land of smallholders and, if so, to document the land-cover changes that followed. The data in Table 2 show that while aggregations occurred in the study region, they do not appear to have involved large, well- capitalized interests. Indeed, the largest new property consisted of seven lots (700 hectares). This finding is consistent with the satellite image analysis, which suggests that the largest of the properties in the region did not expand at the expense of smallholders. Although no participants in the sample were found in the vicinity of the large holdings, deforestation on the individual glebas and the areas surrounding them appeared to be mostly stable and consistent with smallholder activity over the period of interest.

Not only were aggregations relatively infrequent in the sample, countervailing processes of disaggregation and lot fragmentation were also evident. Indeed, the number of disaggregations (16) equaled the number of aggregations (16), although the number of lots that were spun off during disaggregation (25) was slightly less than the number that were absorbed by aggregation (28). Evidently, any trend toward property aggregation is likely to have been diminished by processes of disaggregation, and the structure of the distribution of land holdings by size class in the region does not appear to have moved in the direction of significantly increased social stratification.

Thus, the dynamics of land-cover change over the study period appear to have been associated mainly with autonomous processes that occurred on individual smallholder properties, consistent with life- cycle theories that aim to explain smallholders' farming strategies. The production and land-cover changes associated with annuals (rice, com, and beans) mostly diminished, except for beans in the aggregating and disaggregating property subsamples. Given that beans generally account for little land area and annual production, the overall category effect is a reduced emphasis on annuals independently of property class. The perennials category presents a more complicated picture for several reasons. First, the dynamics of the crops themselves vary by the magnitude of production and area. That is, production may increase while area decreases (e.g., cocoa for stable properties), and, conversely, production may decrease while area increases (e.g., cocoa for disaggregating properties), primarily because of the number of years that these crops require to reach maturity. Second, dynamics across types of crops vary, with coffee generally showing increases in production and area and cocoa and black pepper showing increased area (except on aggregating properties) but decreased production (except on properties at equilibrium).

The dynamics in the cattle sector are strongly consistent with life-cycle theory. The area in pasture increased for all three categories of properties, although somewhat negligibly for those at equilibrium, as did herd sizes. The most dramatic case was for aggregating properties, whose pastures increased by 62 percent over the six-year period (1996-2002), with an average expansion of 10 percent annually. The increase in herd size was even more dramatic, an increment of over 400 percent, or an average of about 67 percent per year.

Conclusion

The panel analysis revealed that during the study period property aggregation was largely compensated for by disaggregation in the study area, and when it occurred, it does not appear to have resulted in the formation of large, well-capitalized ranches. Instead, the aggregation process that did exist arose from the acquisitive interests of the local smallholders themselves and may be regarded as endogenous to the colonization process. In other words, wealthy outsiders do not appear to have penetrated the region and appropriated the land of subsistence farmers. Although these conclusions are based on a panel analysis of a relatively brief duration, they appear to be supported by the satellite images, which show the relative stability of the area cleared by the largest properties in the region. Satellite imagery cannot reveal the actual structure of land ownership, of course, and it is possible that the glebas acquired noncontiguous smallholdings. Our field experience suggests that this is not the case, however, since smallholders in the vicinity have good spatial information about the size of individual fazendas (large farms), which is consistent with the sizes revealed by the images. Nor have we ever encountered a small lot that was owned by a particularly large enterprise.

Thus, the data and analysis are not suggestive of significant frontier stratification, although a skewed distribution of property sizes is plainly evident in the cadastral coverage of Figure 4. Given that the panel analysis shows that aggregation was mostly offset by disaggregation and fragmentation, we conclude that regional patterns of land distribution-and the associated social structure-are largely artifacts of the development decisions made by federal bureaucrats in the 1970s and have not resulted from on-site social dynamics pitting poor farmers against wealthy ranchers. This is not to say that dialectics involving poor farmers and the interests of capital are irrelevant to the Uruar region, since many smallholders migrated there because of technological changes and land consolidation in other parts of Brazil (Hecht 1985). The implication is that the interaction of agents that is of relevance to Uruara's deforestation story has unfolded at the national scale and is not a localized story of emergent social structure.

It is important to note that these results were obtained for a specific time and place. The study period was relatively late in the Amazon development process, which began in the late 1960s, when deforestation on large holdings and ranches probably accounted for most of what was occurring in the region at that time (Yokomizo 1989). Moreover, Uruar is still a remote frontier, where high transportation costs have long constrained economic activity, keeping land rents low. Deforestation processes dominated by smallholders may be expected to take place in similar settings, far from locations of high economic desirability. Elsewhere in the Amazon, the importance of large-scale ranching in land-cover change is likely to be pronounced (Walker, Moran, and Anselin 2000).10

Frontier stratification-with its attendant environmental consequences-could occur over longer time frames than the restricted period of analysis. But our data on property dynamics suggest that disaggregations can mollify productive and speculative aggregations. In fact, this is to be expected with the emergence of a land market and the increasing value of land that follows with the passage of a frontier. Moreover, Thunian models suggest that property sizes decrease with nearness to the market or a reduction in transportation costs, ceteris paribus (Nerlove and Sadka 1991), the beginning of which appears to be occurring in the study area. The analysis presented here calls into question the existence of either a hollow or a speculative frontier along BR-230 near Uruar and perhaps more generally in those parts of the Amazon basin that are undergoing similar processes of colonization.

Evidently, land-cover dynamics in the study region are not attributable to the interactions of heterogeneous agents and emergent social structure, although the aggregating properties do show a marked shift toward ranching, which has implications for land- cover and land-use change in the study area. Such aggregated properties remain relatively small and cannot be confused, under any circumstances, with the large holdings of fazendeiros (largeholder farmers) in the region, who typically possess large ranches covering thousands of hectares.

Thus, the generalized shift toward pasture across property types can be taken as evidence that household-scale dynamics may be the primary driver of the more recent, but nonetheless significant, land- cover changes in the region. However, it appears that th\ese processes are occurring more rapidly than life-cycle theory would suggest. This is not to say that the migration decisions of smallholders did not have an original basis in social processes outside the region, but instead that small and largeholders in Uruar do not present an example of land-use and land-cover change associated with frontier stratification. Although the dynamics of the smallholder farm system that we observed are consistent with life-cycle theory, our demographic analysis suggests that these farm- system changes may not be driven by decreasing dependence and a growing family workforce (Walker 2003), as indicated in Table 5. Resolving the link between the evolution of the farm system and a changing family structure awaits further research.

Loss of the Amazon forest is a critical environmental issue, and much remains to be learned about its underlying causes. This article has addressed forest loss attributable to large- and smallholders and frontier stratification. We found that the farm-level dynamics of smallholders have been the main drivers of the region's deforestation. Consequently, controlling deforestation in frontiers similar to the Transamazon Highway in the vicinity of Uruar will require policies that reshape the agricultural decision making of families, not policies which discourage the acquisitive impulses of well-capitalized ranchers or speculators. This may prove a daunting task, given the large number of smallholders who are active in the Amazon and the Brazilian government's plans to continue colonization projects in the region.

This project was supported by the National Science Foundation (Geography and Regional Science) under the project, "Patterns and Processes of Landscape Change in the Brazilian Amazon (BCS- 0137020)," and hy the National Aeronautics and Space Administration- LBA program under the project, "A Basin-scale Econometric Model for Projecting Future Amazonian Landscapes (NCC5-694)." We are indebted to our tireless field workers and to the residents of Uruara who patiently answered our many questions. The first author thanks Antoinette WinklerPrins and Joseph Messina, who contributed greatly to the master's thesis that served as a rudimentary foundation for this article. Four anonymous reviewers guided us to improve the article substantially, for which we are thankful. Any remaining errors or omissions are solely our responsibility.

1 Large-scale commercial interests may engage in nonranching activities, such as palm oil production, while practically all smallholders have farming systems including cattle and relatively large pastures. Nevertheless, we use the terms largeholder, large- scale commercial interest, and rancher interchangeably. Similarly, we equate smallholder with colonist because this is almost invariably the case in the study area.

2 In the study area, the large landowners, who are simultaneously ranchers, loggers, and land speculators, control the local municipal government and refer to themselves informally as "o senado," or the senate. The most powerful of these individuals arrived as a colonist, worked in a sawmill, and took decades to consolidate his power hase. Thus, the local power structure, which articulates with land distribution, emerged, in large part, through the historic process of colonization; together they constitute a structure of local wealth that represents a significant constraint on newcomers, as well as old timers who lost out in power struggles along the way. The characteristics of structures are highly scale dependent in our usage. For Giddens (1984), of course, agents and structures represent a duality, and in locales, such as the study area, the socially produced nature of the structures (i.e., their "endogenous" nature) is more readily observable than for macroscale structures, which have all the weight and ineluctability of a "natural" force.

3 SUDAM was created as part of "Operation Amazonia" in 1966 and was charged with making the region "attractive for entrepreneurs in other parts of Brazil to reinvest taxable income in the Amazon" (Schmink and Wood 1992, 60).'It has been held responsible by many researchers and activists for environmental and social wrongs, including oversubsidized cattle ranching (Hecht 1985) and the underdevelopment of Amazonian industry (Schmink and Wood 1992). It was replaced by the less powerful Agency for the Development of the Amazon in 2002 after charges of corruption destroyed what remained of SUDAM's credibility.

4 We note the language of Geist and Lambin (2002) and Angelsen and Kaimowitz (1999), who referred to "chain causation" and "logical tandems," whereby sequences of social processes, acts, or behaviors lead to the land-cover outcome.

5 Some researchers (e.g., Maxwell 1999) have also used the term hollow frontier to describe Brazil as a vast area that is devoid of significant European settlement.

6 This is not to say that social equity is not a valid topic of study; we mean to point out only that the environmental implications of frontier stratification have not been well studied.

7 This is difficult to assess, however, because MSS (Landsat Multi-Spectral Scanner) imagery does not appear to be available for this period in this location.

8 It is possible that large ranches possess noncontiguous lots, in which case the deforestation for the glebas could be underestimated to a certain degree. We have not encountered many such disaggregate holdings, however. In addition, the aggregation analysis suggests that such forces of concentration are quiescent in the region.

9 1996 is the first year that the Gini coefficient can be measured in the study area, given the municipio's founding in that year. Perfect equality results in a Gini coefficient of 0.

10 Soy is becoming increasingly important, but to date its expansion is found mostly in cerrado and in abandoned pastures, in which case the impacts on deforestation have so far remained low.

References

Alston, L. J. 1996. The determinants and impact of property rights: Land titles on the Brazilian frontier. Journal of Law, Economics & Organization 12:25-61.

Alston, L. J.; Libecap, G. D.; and Mueller, B. 2000. Land reform policies, the sources of violent conflict, and implications for deforestation in the Brazilian Amazon. Journal of Environmental Economics and Management 39:162-88.

Alston, L. J.; Libecap, G. D.; and Schneider, R. 1993. The settlement process, property rights, and land use on the Brazilian agricultural irontier: Preliminary results and lessons from U.S. economic history. Paper presented at the All-University of California Economic History Conference, Pasadena.

Anderson, L. E.; Granger, C. W. J.; Reis, E.; Weinhold, D.; and Wunder, S. 2002. The dynamics of deforestation and economic growth in the Brazilian Amazon. Cambridge, U.K.: Cambridge University Press.

Angelsen, A., and Kaimowitz, D. 1999. Rethinking the causes of deforestation: Lessons from economic models. The World Bank Research Observer 14(1):73-98.

Arima, E. Y., and Uhl, C. 1997. Ranching in the Brazilian Amazon in a national context: Economics, policy, and practice. Society and Natural Resources 10:433-51.

Arima, E.; Walker, R. T.; Perz, S. G.; and Caldas, M. 2005. Loggers and forest fragmentation: Behavioral models of road building in the Amazon basin. Annals of the Association of American Geographers 95:525-41.

Binswanger, H. 1991. Brazilian policies that encourage deforestation in the Amazon. World Development 19:821-29.

Blaikie, P. and Brookfield, H. 1987. Land degradation and society. London: Methuen.

Brondzio, E. S. 1997. From extractivists to forest farmers: Changing concepts of caboclo agroforestry in the Amazon Estuary. Research in Economic Anthropology 18:233-79.

Browder, J. O. 1988. Public policy and deforestation in the Brazilian Amazon. In Public policies and the misuse of forest resources, edited by R. Repetto, 247-97. Cambridge, U.K.: Cambridge University Press, for World Resources Institute.

Brown, J. C.; Jepson, W.; and Price, K. P. 2004. Expansion of mechanized agriculture and landcover change in southern Rondnia, Brazil. Journal of Latin American Geography 3:96-102.

Brown, L. A.; Sierra, R.; and Digiacinto, S. 1994. Urban-system evolution in frontier settings. Geographical Review 84:249-65.

Bryant, R. L. 2001. Political ecology: A critical agenda for change? In Social nature: Theory, practice and politics, edited by N. Castree and B. Braun, 151-69. Malden, Mass.: Blackwell.

Bunker, S. G. 1981. The impact of deforestation on peasant communities in the medio Amazonas of Brazil. In Where have all the flowers gone? Deforestation in the Third World, ed. V. H. Sutlive, N. Altshuler, and M. Zamora, 45-60. Williamsburg, Va.: Department of Anthropology, College of William and Mary.

Casetti, E., and Gauthier, H. L. 1977. A formalization and test of the "hollow frontier" hypothesis. Economic Geography 53:70-78.

Censo-Agropecurio. 1995-1996. Brazil: Censo Agropecurio.

Coomes, O. T., and Burt, G. J. 2001. Peasant charcoal production in the Peruvian Amazon: Rainforest use and economic reliance. Forest Ecology and Management 140(1):39-50.

da Veiga, J. J.; Tourrarid, J.; and Quanz, D. 1996. A pecuria na fronteira agrcola da Amaznia: O caso do municpio de Uruar, PA, na Regio da Transamaznica [Cattle on the agricultural frontier of Amazonia: The case of Uruar county, Para in the Transamazon region]. EMBRAPA/CPATU.

Faminow, M. D. 1998. Cattle, deforestation and development in the Amazon: An economic, agronomic and environmental perspective. 1st ed. New York: CAB International.

Fearnside, P. M. 1986. Spatial concentration of deforestation in the Brazilian Amazon. Ambio 15(2):74-81.

_____. 1993. Deforestation in Brazilian Amazonia: The effect of population and land tenure. Ambio 22:537-45.

Foweraker, J. 1981. The struggle for land: A political economy of the pioneer frontier in Brazil from 1930 to the present day, e\d. M. Deas. Cambridge, U.K.: Cambridge University Press.

Geist, H. J., and Lambin, E. F. 2002. Proximate causes and underlying driving forces of tropical deforestation. BioScience 52:143-50.

Giddens, A. 1984. The constitution of society: Outline of the theory of structuration. Berkeley: University of California Press.

Godoy, R.; Jacobson, M.; de Castro, J.; Aliaga, V.; Romero, J.; and Davis, A. 1998. The role of tenure security and private time preference in neotropical deforestation. Land Economics 74:162-70.

Godoy, R.; Wilkie, D.; and Franks, J. 1997. The effects of markets on neotropical deforestation: A comparative study of four Amerindian societies. Current Anthropology 38:875-78.

Gutman, G.; Janetos, A.; Justice, C.; Moran, E. F.; Mustard, J.; Rindfuss, R. R.; Skole, D. L.; Turner, B. L., II; and Cochrane, M. A. 2004. Land change science: Observing, monitoring, and understanding trajectories of change on the Earth's surface. Vol. 6, Remote sensing and digital image processing. New York: Kluwer Academic.

Hall, A. L. 1989. Developing Amazonia: Deforestation and social conflict in Brazil's Carajas Programme. Manchester, U.K.: Manchester University Press.

Hecht, S. B. 1984. Cattle ranching in Amazonia: Political and ecological considerations. In Frontier expansion in Amazonia, ed. M. Schmink and C. H. Wood, 366-98. Gainesville: University of Florida Press.

_____. 1985. Environment, development and politics: Capital accumulation and the livestock sector in eastern Amazonia. World Development 13:663-84.

_____. 1993. The logic of livestock and deforestation in Amazonia: Considering land markets, value of ancillaries, the larger macroeconomic context, and individual economic strategies. BioScience 43(10):687-95.

Hecht, S. B., and Anderson, A. B. 1988. The subsidy from nature: Shifting cultivation, successional palm forests, and rural development. Human Organization 47:25-35.

Hecht, S., and Cockburn, A. 1990. The fate of the forest: Developers, destroyers, and defenders of the Amazon. 1st ed. New York: HarperCollins.

INPE. 2005. Monitoriarnento da Floresta Amaznia Brasileira par satlite: Projecto Prodes [The monitoring of the Brazilian Amazon forest by satellite: Project Prodes], ed. I. N. d. P. Espacias: Governo do Brasil.

James, P. E. 1969. Latin America. 3rd ed. New York: Odyssey Press.

Jepson, W. 2005. A disappearing biome? Reconsidering land-cover change in the Brazilian savanna. Geographical Journal 171:99-111.

Mahar, D. J. 1979. Frontier development policy in Brazil: A study of Amazonia, New York: Praeger.

_____. 1989a. Government policies and deforestation in the Brazilian Amazon. In Environmental management and economic development, ed. G. Schramm and J. J. Warford, 87-116. Baltimore, Md.: Johns Hopkins University Press.

_____. 1989b. Government policies and deforestation in Brazil's Amazon region. Washington, D.C.: World Bank.

Margulis, S. 2004. Causes of deforestation of the Brazilian Amazon. Washington, D.C.: World Bank.

Maxwell, K. 1999. The two Brazils. The Wilson Quarterly, Winter, 50-60.

McCracken, S. D.; Brondizio, E.; Nelson, D.; Moran, E. F.; Siqueira, A.; and Rodriguez-Pedraza, C. 1999. Remote sensing and GIS at farm property level: Demography and deforestation in the Brazilian Amazon. Photogrammetric Engineering & Remote Sensing 65(11):1311-20.

Messina, J. P., and Walsh, S. J. 2000. The application of a cellular automaton model for predicting deforestation: Patterns and processes of LULCC in the Ecuadorian Amazon. Paper presented at the Fourth International Conference on Integrating GIS and Environmental Modeling (GIS/EM4): Problems, Prospects and Research Needs, 2-8 September, Banff, Alberta, Canada.

Moran, E. F. 1984. Colonization in the Transamazon and Rondnia. In Frontier expansion in Amazonia., ed. M. Schmink and C. H. Wood, 285-303. Gainesville: University of Florida Press.

_____. 1989. Government-directed settlement in the 1970s: An assessment of Transamazon highway colonization. In The human ecology of tropical land settlement in Latin America, ed. D. A. Scumann and W. L. Partridge, 172-98. Boulder, Colo.: Westview Press.

Myers, N. 1993. Tropical forests: The main deforestation fronts. Environmental Conservation 20:9-16.

Nehmi Filho, V. A. N., and Pusch, J. 2003. Avanos dos gros nas terras de pastagens do Centro-Oeste [The advance of grain on the pasture land of the center-west]. Annualpec, 333-36. So Paulo: FNP Consultoria.

Nepstad, D.; Coapobianco, J. P.; Barros, A. C.; Carvalho, G.; Moutinho, P.; Lopes, U.; and Lefebvre, P. 2002. Roads in the rainforest: Environmental costs for the Amazon. Belm, Brazil: Amazon Institute of Environmental Research (IPAM, Brazil).

Nerlove, M., and Sadka, E. 1991. VonThnen's model of the dual economy. Journal of Economics 54:97-123.

Ozorio de Almeida, A. L., and Campari, J. S. 1995. Sustainable settlement in the Brazilian Amazon. Oxford, U.K.: Oxford University Press.

Perz, S. G. 2002. Population growth and net migration in the Brazilian Legal Amazon, 1970-1996. In Deforestation and land use in the Amazon, edited by C. H. Wood and R. Porro, 107-129. Gainesville: University Press of Florida.

Pfaff, A. S. P. 1999. What drives deforestation in the Brazilian Amazon? Journal of Environmental Economics and Management 37(1):26- 43.

Pichn, F. J. 1997. Settler households and land-use patterns in the Amazon frontier: Farm-level evidence from Ecuador. World Development 25(1):67-91.

Porro, R. 2002. Land use, cattle ranching and the concentration of landownership in Maranho, Brazil. In Deforestation and land use in the Amazon, ed. C. H. Wood and R. Porro, 315-37. Gainesville: University Press of Florida.

Reis, E., and Guzmn, R. 1994. An economic model of Amazon deforestation. In The causes of tropical deforestation: The economic and statistical analysis of factors giving rise to the loss of tropical forests, ed. K. Brown and D. Pearce, 172-92. London: University College of London Press.

Robbins, P. 2004. Political ecology. Critical introductions to geography. Malden, Mass.: Blackwell.

Roberts, J. T., and Thano

Source: Economic Geography

More News in this Category