Amazonia and Global Change (Geophysical Monograph)

"

Geophysical Monograph 186

Amazonia and Global Change Michael Keller Mercedes Bustamante John Gash Pedro Silva Dias

Editors

~ American Geophysical Union Washington, DC

• • 25 ha. Small deforestation events «25 ha, not shown) in Mato Grosso account for 15% of all deforestation. Percentages refer to the fraction of cleared area converted to specific land uses [Morton et al., 2006, 2007a].

ALVES ET AL.

16 DEFORESTAnON AND LAND USE IN BRAZILIAN AMAZONIA

17

i

Table 1. Evolution ofAggregate Land Use Statistics According to Brazilian Agricultural Census· 1970

1975

1980

1985

1995

Land Category, % Total Farm Area

Pasture Crops, temporary Crops, permanent Forest Abandoned landb All otherC

37.9 2.6 0.3 37.3 15.5 6.4

35.5 3.3 0.4 43.0 13.1 4.7

35.2 4.1 0.7 42.4 9.8 7.8

36.8 4.4 0.8 40.4 8.9 8.7

42.4 3.9 0.8 41.3 5.7 5.9

Temporary Crops in MT and in All Other States, % Total Farm Area

Mato Grosso state All other states

1.4 3.0

2.1 3.8

4.1 4.0

5.3 4.0

5.6 2.8

Cattle Head in Amazonia as a Fraction a/National Herd, %

8.2

9.0

12.7

14.7

23.3

Average Stocking Rate, headlha

0.30

0.30

0.40

0.40

0.70

·Aggregate data for the nine states belonging to Legal Amazonia: Acre, Amapa, Amazonas, Maranhao, Mato Grosso, Para, Rondonia, Roraima, and Tocantins. Source: http:// www.ipeadata.gov.br. bAbandoned land is defined as land unused for more than 4 years. cAll other includes land in rotation, planted forest, and other categories like swamps.

farms; cropland can rotate with pasture when grain prices are low. Although not shown in Figure 2, agroforestry, reforestation, urban expansion, and other types ofland use can also replace pastures or croplands. In addition, some land use trajectories can be influenced by a combination of factors, such as forest degradation from selective logging and fire [Nepstad et al., 1999], which fundamentally alter forest structure and land value. Particular LCLUC transitions generate unique patterns of forest loss. Different agrarian regimes, inc~uding farm size, the architecture of settlement projects, and different production and land management strategies can lead to diverse expressions of the same trajectory in landscape patterns. The composition and configuration of the landscapes produced have important consequences for the functioning of the biophysical systems in Amazonia and may help inform discussions of plausible development scenarios for the region. Within many older agricultural frontiers, concentrated deforestation activity in the vicinity of major roads and colonization projects [Machado, 1998; Alves, 2002] has led to landscapes dominated by pastures and cropland. The magnitude of forest clearing for agriculture in these areas often exceeds the limits prescribed by the Brazilian Forest Code [Alves et al., 2003; Alvez, 2007b].

The following sections review advances in understanding the evolution oflandscape patterns and the dominant longterm LCLUC trajectories in Amazonia.

eastern Para states and in smaller clearings in regions with higher densities of settlement projects in Para and Rondonia. Overall, large e-(earings on larger farms contributed the greatest fraction)bf total deforestation (Figure 3) [Alves, 2002]. During ~00-2005, the patterns in deforestation size show a bimodal distribution, with regions either dominated by very large (> 1000 ha, 25% of cells) or very small «50 ha, 51 % of cells) clearing sizes. Very large clearings in central Para, southern Amazonas, and central Roraima states suggest that these regions were recently exposed to the same degree of capital and technology that was previously found only in older frontier areas. Landscape patterns of forest conversion at the local scale reflect additional heterogeneity beyond clearing size (Plate 3). In 1986, central Rondonia near Jaru was already highly disturbed, consisting of nearly equal proportions of primary forest and pasture, with pasture concentrated along planned roads at 4-km intervals. By 2003, linear strips of forest from 1986 had been reduced to small forest fragments, many of which were less than 1 km across. Small patches of secondary forest mapped in 2003 occur exclusively along the margins offorest fragments that have never b~en cleared, suggesting that forest edges are taking on the spectral signature of secondary forest in the absence of any clearing. Patterns in a region of the nearby municipio of Ariquemes differ markedly with extensive tracts ofmature forest in both 1986 and 2003, no "fishbone" pattern from evenly spaced roads, and some large patches of secondary forest as much as 18 years old. Differences in fragmentation patterns reflect differences in the architecture of settlement and colonization

projects, whereas the persistence of secondary forest in the northwest is likely due to higher rainfall and poorer soils in this region. 3.2. Forest Degradation From Logging Selective logging is one of the most important drivers of forest degradation and land cover change in Amazonia. Logging is rarely practiced in a sustainable fashion. In fact, only 1248 ha of mature forests were harvested following the Forest Stewardship Council (FSC) standards in Amazonia in 2003 [Lentini et al., 2005]. The extensive network of secondary roads built by loggers and capital obtained by land owners selling timber help to accelerate the deforestation process near sawmill centers [Uhl et al., 1991; Verissimo et al., 1992]. Unmanaged logging practices lead to forest degradation through damage to forest structure and altered species composition [see Asner et al., this volume]. Using remote sensing techniques, Asner et al. [2005] estimated that the annual area affected by logging was 12,000-19,000 km2 between 2000 and 2002, equivalent to the average annual deforestation rate during this period of 18,000 ± 2900 km [INPE, 2007]. Logging and deforestation are not mutually exclusive; an average of 16% oflogged forests ¥e clear-cut in the first year following logging operations, with 33% deforested within 4 years of logging [Asner et al;, 2006]. Canopy damage and slash from logging operatiods increase the likelihood of fire damage in logged forests [Nepstad et al., 1999], although the extent of logged and burned forest has not been estimated for the entire Amazon region.

3.1. Landscape Patterns ofForest Conversion Deforestation in Amazonia has replaced the forest with a fragmented landscape ofpasture and agricultural areas, leaving few forest remnants where deforestation has been most concentrated. The total extent of deforestation in Amazonia until 2005, depicted in Plate 1, provides a first approximation of important regional patterns in forest loss. Major road and river networks are buffered by the outlines of historic deforestation and older frontier areas of eastern Para, Mato Grosso, and Rondonia states have greater forest loss than newer frontiers in central Para, Acre, or Amazonas states. Specific site conditions, including soil quality or topography, further influence both the location of forest clearing and the postclearing land uses, such that patterns of deforestation and land use may be locally consistent. The spatial patterns resulting from forest conversion may differ substantially across the basin as a function of clearing size (Plate 2). Deforestation between 1991 and 1997 occurred in very large clearings in central Mato Grosso and

,-.. ~

3.3. Forest Conversion to Pasture

100

= '-" ~

e

~

75

~

e

I-

~

.... .........

50

•••••••••••••••

~ ~

Q

l:l

~OJ

e

""

0 0

25 50 75 Number of cells (%)

100

!-1991-97mte • area200ha!

Figure 3. Lorenz curve of the 1991-1997 deforestation rate calculated for \1..0 cells and accompanying cumulative curves of the area of forest clearings of two different sizes in the same period [after Alves, 2002].

According to census data, pasture has been the most common land use in Amazonia (Table 1). Typical processes of pasture establishment in Amazonia include the direct conversion of forest to pasture or a longer conversion trajectory beginning with an initial phase of annual crops before pasture establishment after a number ofyears [e.g., Millikan, 1992]. After establishment, pasture productivity typically remains high for 5 to 7 years before declining due to changes in soil fertility and pH, resulting in a progressive decrease in forage quality and increased weed invasion [Buschbacher, 1986; Serrao and Toledo, 1990]. As pasture quality degrades, pastures can either be reinvigorated through repeated cycles ofburning, reformed via fertilizer application and reseeding, or abandoned to secondary succession. The length oftime a pasture remains productive is highly dependent on pasture management practices, local climate, and soil quality [Serrao and Toledo, 1990;

18

DEFORESTATION AND LAND USE IN BRAZILIAN AMAZONIA

Moran, 1993; Dias Filho et al., 2000; Numata et al., 2007]. For example, pastures established on Alfisols or Ultiso1s in Rondonia that receive moderate levels of precipitation can remain prdtluctive for well over 20 years, whereas pastures established on Oxisols or in more humid or arid conditions show earlier evidence of degradation and higher rates of abandonment [Numata et ai., 2007]. Cattle ranching remains the dominant land use in Amazonia (Table 1), following important changes during the last decades. Faminow [1998] argues that a fundamental cause for the growth of the cattle herd was the considerable expansion of regional demand associated with urban growth. Andersen et ai. [2002] and Margulis [2004] reviewed the many motivations for cattle ranching and intensification of pasture use, concluding that ranching became profitable independent of subsidies due to the growth of urban demand and increased productivity. Veiga et al. [2004] observed a variety of market chains stimulated by local demands and markets outside Amazonia. Higher stocking rates are more commonly found in the most deforested areas, suggesting a transition to pasture use intensification [Alves, 2007a]. Taken together, these factors help to explain the continued predominance of pastures in land use trajectories in Amazonian landscapes. 3.4. Forest Conversion to Cropland In the context of LCLUC, we classify forest conversion to cropland according to the most common land use trajectories in recent decades. Cropland may directly follow deforestation or arise as part of a rotation cycle with secondary forest or pasture.' Direct conversion of forest to cropland occurs for both small-scale [e.g., Moran and Brondizio, 1998] and large-scale crop production [Morton et al., 2006]. In addition to subsistence crops, small farmers may also invest in other crops for local or national markets [Moran and Brondizio, 1998; Costa, 2007]. Forest conversion for soybean, maize, or other grain production follows the recent development of crop varieties specifically adapted to the soils and climate of some Amazon regions [Warnken, 1999; Jepson, 2006]. The dynamics of forest conversion for mechanized crop production in Mato Grosso is discussed in more detail in section 4.1. The nature ofrotation systems between cropland and forest or pasture depend on both farm size and market conditions. For small farms, crop areas may be used until soil nutrients are depleted and then abandoned for several years to allow forest vegetation to accumulate nutrients. The length of fallow rotations in a "slash-and-burn" or "chop-and-mulch" system depends on the rate of forest recovery and farm size [Denich et al., 2004]. On larger farms, market conditions

for beef or grains may determine the interannual patterns of pasture and cropland use or the frequency of fallow cycles. Cropland can be both a precursor to land consolidation for cattle ranching or an endpoint itself, independent offarm size. Census data suggest that in Amazonia, croplands established in the original phases of colonization were largely replaced by cattle ranching as more forest was converted [Alves, 2007a]. However, recent expansion of mechanized crop production was generated through new deforestation, savanna clearing, and transitions from pasture to cropland [Morton et al., 2006]. The diversity of transition pathways, crop types, and farm sizes in Amazonia highlights the spatial and temporal variability of cropland on the landscape. Deforestation dynamics in Mato Grosso state represents one case ofparticular interest because of specific sociodemographic, economic, and bioclimatic conditions, suggesting the establishment of a new land use system differing from those dominating in other parts of Amazonia. Mato Grosso had the highest deforestation rate during 1995-2005, accounting for 33-43% of the annual deforestation increment in the Brazilian Amazon [lNPE, 2007]. High rates of forest loss were driven by large clearing sizes [Alves, 2002; Morton et al., 2006; Ferreira et ai., 2007] (Plate 2 and Figure 3); large landholders (2:1000 ha) owned an estimated 84% and 82% of all land in private property statewide according to the 1985 and 1996 agricultural censuses, respectively [lBGE, 1996]. Although deforestation is associated with a variety of influences, economic factors have been largely linked with credit and economic opportunities for extensive cattle ranching operations and crop production such as soybeans, and with inter-regional differences in land prices [Fearnside, 2001; Andersen et al., 2002; Margulis, 2004; Morton et ai., 2006]. Deforestation in Mato Grosso is highly mechanized in comparison with other states. Two tractors, linked by a strong chain, are used to pull down trees in the transitional forests. Even in taller-stature forests, heavy machinery is used to manage manually felled trees. Piling and re-burning fore~t vegetation can reduce standing forest to bare soil in a matter of months. Unlike previous estimates of carbon losses from deforestation, where 20% of biomass is combusted, and the remainder decomposes over 10-30 years [Fearnside et al., 1993; Houghton et ai., 2000], mechanized forest clearing practices may result in near-complete combustion of aboveground woody biomass and woody roots [Morton et al., 2008]. Mechanization has the~eby increased the potential size of forest clearings and decreased the duration of the deforestation process. Combined advances in deforestation mapping and tracking the fate of cleared land provide spatial and temporal details regarding land cover transitions statewide. Vegetation

ALVES ET AL.

phenology, derived froriI time series of MODIS data, has proven useful for separating land cover types and following changes in land mauagement over time [Ratana et al., 2005; Morton et ai., 20.d6; Brown et al., 2007]. Figure 2b highlights the dyna~cs of 2000-2005 transitions among major cover types in'Mato Grosso state, showing the proportion of new deforestation, woodland savanna, and secondary forest conversions >25"ha as a function of postclearing land use. The main driver of forest loss in Mato Grosso is large-scale cattle production, yet direct conversion of forest to cropland contributed substantially to the number of large deforestation events and to woodland and secondary forest losses during this period [Morton et al., 2006, 2007a, 2007b]. Secondary forest is not a large component of the landscape in Mato Grosso compared with estimates for other regions, comprising only 11-14% of historic deforestation [Carreiras et ai., 2006; Morton et al., 2007a]. Detailed analysis of the source of secondary and degraded forests in Mato Grosso from abandonment, logging, and burning remains a research challenge. Expansion of soybeans and other mechanized crop varieties in Amazonia has renewed the debate over extensive versus intensive land uses, and about the social and environmental outcomes of agricultural expansion. Climate, soils, and topography are suitable for soybean cultivation in forested regions of northern Mato Grosso and surrounding areas [Jasinski et ai., 2005], and some authors have argued that soybean cultivation can be a competitive, intensive agriculture alternative over extensive and low-productive cattle ranching [e.g., Andersen et al., 2002; Margulis, 2004]. However, soybean production can contribute to pushing cattle ranching into new deforestation frontiers, as seen following its introduction in southern and west-central Brazil [Andersen et al., 2002], even if a detailed assessment of the role of soybeans in concentration of land tenure and income, rural outmigration, and loss of biodiversity has not yet been completed [Fearnside, 2001]. 3.5. Land Abandonment and Secondary Vegetation Growth

Considerable research has focused on mapping secondary forest at local and regional scales [Lucas et al., 1993; Moran et ai., 1994; Roberts et ai., 2002; Alves et al., 2003; Carreiras et ai., 2006]. Secondary forests are a potential carbon sink and can help recover hydrological and biogeochemical functioning after forest clearing [e.g., Brown and Lugo, 1990; Moran et al., 1994]. Secondary succession can develop following different pathways, including land rota~ tion during shifting cultivation and land abandonment after pasture degradation or immediately following forest clearing; species composition, vegetation structure, and rates of carbon uptake in secondary forests are highly dependent

19

upon soil type and prior land use [Alves et al., 1997; Moran et al., 2000; Lucas et ai., 2002; Zarin et al., 2005]. Census data and remote sensing analyses raise important questions about the long-term dynamics of secondary vegetation in Amazonia. The proportion of cleared land that was unused for more than 4 years as a percentage of farm area declined steadily, from 15.5% to 5.7%, during 1970-1995 (Table 1). This evidence is consistent with findings that rates of land abandonment were higher in newly established frontiers, while secondary vegetation tended to be re-cleared concurrently with the elimination of mature forest remnants in older settlement areas [Alves et al., 2003; Alves, 2007a]. Time series of satellite data show that secondary forest is a dynamic component of the landscape in the Ariquemes and Ii-Parana regions of Rondonia (Figure 4). In both regions, steady increases in pasture area resulted from more rapid re-clearing of secondary forest than pasture abandonment. Overall, the contribution of secondary forest remained stable or declined during 1986-2003, never exceeding 10% of the landscape. Declining rates of land abandonment in more intensively deforested areas indicate that over the long term, secondary forests may offset only a small fraction of the initial carbon emissions from deforestation [Alves et ai., 1997; Alves,2007a]. 4. cmJcLusIONS AND OUTLOOK ,

Brazilian Amaionia is one of most active regions of agricultural expansipn in the world. Clearing tropical forest is the primary means to increase the area of cattle pasture and crops, while related processes such as logging, fire for land clearing and management, land abandonment, and land use intensification are also key elements of the LCLUC dynamics. The conceptual model of transitions between multiple land cover and use states' illustrates the heterogeneity of LCLUC trajectories and their expression in landscape patterns across Amazonia. Characterizing the spatial patterns created by such processes represents an important methodological success in Amazonia, based on multiple data sources and a variety of analysis techniques, from which to investigate the role of LCLUC on the biophysical system. Agriculture in the region is becoming increasingly intensive, conducted by large-scale operators with sufficient access to capital. These shifts in the spatial and temporal dynamics of LCLUC are present in both census data and satellite remote sensing as a decrease in secondary forests, increase in pasture stocking rates, and rapid expansion of the area under mechanized agriculture. The rise of intensive production and the influence of both national and international market forces on land use have led to the development of new ecologically oriented certification schemes for beef

20

DEFORESTATION AND LAND USE IN BRAZILIAN AMAZONIA

A) Ji-Parana

Pasture

ALVES ET AL.

B) Ariquemes

311

40 315

30

30

211

25

20

20

111

115 10

10

1:

II

'~ofl1t+~HHft+t1~~Hilff "

6.

5

II

10

20

.\

101.---------------1

.

C} JI-Parana

D Ari uemes

~...;;.;;..;;;..;;;;;.=~------.....,

10

2Or-------.::;;..L.;=-=;;.;;L;=~

..

1:

l:

'"

..'"

'"~

in Brazilian Amazonia provides detailed estimates of forest loss on an annual basis. Advancement in near real-time monitoring of defor~tationin cerrado and closed forest and mapping selective,t10gging has generated essential data for environmental ~nitoring. Successes in remote sensing of deforestation in Amazonia serve as an important example of technical progress for other nations considering programs to reduce deforestation. Future research will continue to focus on the economic, social, and environmental elements of each forest loss trajectory, highlighting spatial and temporal heterogeneity in the causes and consequences of Amazon deforestation. Recent advances in remote sensing pave the way for additional efforts to quantify the basin-wide impacts offorest degradation from fire, forest fragmentation, and land abandonment to secondary forest. Findings from LBA also lay the groundwork for related research on the influence of specific land use transitions and spatial patterns of land cover for climate, biogeochemistry, and long-term agricultural productivity, as reported in the following chapters of this book.

10

REFERENCES

~

6.

0

r.

0

10

10

Secondary Forest Figure 4. Transitions among pasture (gray), secondary forest (dashed), and primary forest (black) for the Ii-Parana and Ariquemes regions of Rondonia state during 1986-2003. Dynamics for pasture in (a) Ji-Parana and (b) Ariquemes. Values above the x axis represent a gain of a specific class as a percentage of the landscape, and values below the axis represent a loss. Changes in secondary forest 'over time in a similar manner for (c) Ii-Parana and (d) Ariquemes. In Ji-Parana, pasture shows a general increase over time, with most pasture originating from areas that were previously pasture. Pasture loss is primarily to secondary forest. Secondary forest shows no significant increase over time in Ji-Parana, leading to a declining ratio of secondary forest to cleared lands. Large fluctuations between pasture and secondary forest in Ii-Parana during 1997-1999 are most likely due to early dry season imagery in these years leading to overestimating secondary forest. Rates of pasture abandonment to secondary forest were more stable in Ariquemes than in Ji-Parana. Both pasture and secondary forest show a general increase over time, resulting in a ratio of secondary forest to pasture of over 30% in Ariquemes.

and grain production in Amazonia. At the same time, high deforestation rates in older settlement areas, expansion of agricultural frontiers into new areas, and prevailingly low productivity ofland show the recurrence of historical trends. Thus, a diversity of actors remain influential in both "old"

and "new" frontiers presenting a challenge for delineating plausible future scenarios ofLCLUC in Amazonia. Advances in satellite remote sensing of deforestation and postclearing land use have led to high-quality data for both science and policy applications. Deforestation mapping

Alves, D. S. (2002), Space-time dynamics of deforestation in Brazilian Amazonia, Int. J. Remote Sens., 23, 2903-2908. Alves, D. S. (2007a), Cenarios de cobertura e uso da terra e dimensoes humanas no LBA, in Dimensoes Humanas da BiosferaAtmosfera daAmazonia, edited by W. M. da Costa, B. K. Becker, and D. S. Alves, pp. 39-63, EDUSP, Sao Paulo, Brazil. Alves, D. S. (2007b), Science and technology and sustainable development in Brazilian Amazon, in The Stability of Tropical Rainforest Margins, edited by T. Tscharntke et aI., pp. 493-512, Springer, Berlin, Germany. Alves, D. S., J. V. Soares, S. Amaral, E. M. K. Mello, S. A. S. Almeida, O. F. da Silva, and A. M. Silveira (1997), Biomass of primary and secondary vegetation in Rondonia, Western Brazilian Amazon, Global Change BioI., 3, 451--461. Alves, D. S., M. 1. S. Escada, J. L. G. Pereira, and C. A. Linhares (2003), Land use intensification and abandonment in Rondonia, Brazilian Amazonia, Int. J. Remote Sens., 24,899-903. Andersen, L. E., C. W. J. Granger, E. J. Reis, D. Weinhold, and S. Wunder (2002), The Economics of Deforestation: Dynamic Modeling ofAmazonia, Cambridge Univ. Press, Cambridge. Anderson, L. 0., Y. E. Shimabukuro, R. S. DeFries, and D. C. Morton (2005), Assessment of deforestation in near real time over the Brazilian Amazon using multitemporal fraction images derived from Terra MODIS, IEEE Geosci. Remote Sens. Lett., 2, 315-318. Asner, G.P. (2001), Cloud cover in Landsat observations of the Brazilian Amazon, Int. J. Remote Sens., 22, 3855-3862. Asner, G. P., D. E. Knapp, E. N. Broadbent, P. J. C. Oliveira, M. Keller, and J. N. Silva (2005), Selective logging in the Brazilian Amazon,Science,310,480--482.

21

Asner, G. P., E. N. Broadbent, P. J. C. Oliveira, M. Keller, D. E. Knapp, and J. N. M. Silva (2006), Condition and fate of logged forests in the Brazilian Amazon, Proc. Natl. Acad. Sci. U. S. A., 103,12,947-12,950, doi:1O.1073/pnas.0604093103. Asner, G. P., M. Keller, M. Lentini, F. Merry, and C. Souza Jr. (2009), Selective logging and its relation to deforestation, Geophys. Monogr. Ser., doi:lO.l029/2008GM000723, this volume. Batistella, M., and E. F. Moran (2005), Dimensoes humanas do uso e cobertura das terras na Amazonia: Uma contribuifYao do LBA, Acta Amazonica, 35,239-247. Batistella, M., S. Robeson, and E. F. Moran (2003), Settlement design, forest fragmentation, and landscape change in Rondonia, Amazonia, Photogramm. Eng. Remote Sens., 69, 805-812. Becker, B. K. (1997), Amazonia, 5th ed., ATICA, Sao Paulo. Brown, J. C., W. E. Jepson, J. H. Kastens, B. D. Wardlow, J. M. Lomas, and K. P. Price (2007), Multitemporal, moderate-spatial resolution remote sensing of modem agricultural production and land modification in the Brazilian Amazon, GIScience Remote Sens., 44,117-148. Brown, S., and A. Lugo (1990), Tropical secondary forests, J. Tropical Ecol., 6, 1-32. Buschbacher, R. (1986), Tropical deforestation and pasture development, Bioscience, 36,22-28. Cardille, J. A., and J. A. Foley (2003), Agricultural land-use change in Brazilian Amazonia between 1980 and 1995: Evidence from integrated satellite and census data, Remote Sens. Environ., 87, 551-562. Carreiras, J. M. B., y(. E. Shimabukuro, and J. M. C. Pereira (2002), Fraction images derived from SPOT-4 VEGETATION data to assess land-cover Ichange over the State ofMato Grosso, Brazil, Int. J. Remote Sen's., 23, 4979--4983. Carreiras, J. M. B., J. M. C. Pereira, M. L. Campagnolo, and Y. E. Shimabukuro (2006), Assessing the extent of agriculture/pasture and secondary succession forest in the Brazilian Legal Amazon using SPOT VEGETATION data, Remote Sens. Environ., 101, 283-298. Chambers, J. Q., G. P. Asner, D. C. Morton, L. O. Anderson, S. S. Saatchi, F. d. B. Espirito-'Santo, M. Palace, and C. Souza Jr. (2007), Regional ecosystem structure and function: Ecological insights from remote sensing of tropical forests, Trends Ecol. Evoi,22,414--423. Costa, W. M. (2007), Tendencias recentes na Amazonia: Os sistemas produtivos emergentes, in Dimensoes Humanas da Biosfera-Atmosfera da Amazonia, edited by W. M. da Costa, B. K. Becker, and D. S. Alves, pp. 81-11, EDUSP, Sao Paulo, Brazil. Denich, M., K. Vielhauer, M. S. de A. Kato, A. Block, O. R. Kato, T. D. de Abreu Sa, W. Lucke, and P. L. G. Vlek (2004), Mechanized land preparation in forest-based fallow systems: The experience from eastern Amazonia, Agroforestry Syst., 61, 91-106. Dias Filho, M. B., E. A. Davidson, and C. J. R. de Carvalho (2000), Linking biogeochemical cycles to cattle pasture management and sustainability in the Amazon Basin, in The Biogeochemistry ?fthe Amazon Basin, edited by M. McClain, R. L. Victoria, and J. E. Ritchey, pp. 85-105, Oxford Univ. Press, New York. Faminow, M. D. (1998), Cattle Deforestation and Development in the Amazon, CAB International, New York.

ALYES ET AL. 22

23

DEFORESTATION AND LAND USE IN BRAZILIAN AMAZONIA

Honzak G. M. Foody, P. 1. Curran, and C. Cor) , . Lucas, R . M 0" M , (1990) , The rate and extent of deforestation in F earnsl'd e, P .M ves (1993), Characterising tropical secondary forests usmg . Brazilian Amazonia, Environ. Conserv., 17, 213-226. multi-temporal Landsat sensor imagery, Int. J. Remote Sens., 14, Feaffii~de, P. M. (2001), Soybean cultivation as a threat to the envI3061-3067. ronment in Brazil, Environ. Conserv., 28, 23-38. Lucas, R. M., M. Honzak, I. do Amaral, P. J. Curran, and G. Jy,1. 'd P MAT Tardin and L. G. Meira Filho .(1990), F earnsl e, . ., . . , Foody (2002), Forest regeneration on abandoned clearances m Deforestation rate in Brazilian Amazonia, Reprint, InstItuto de Central Amazonia, Int. 1. Remote Sens., 23, 965-9~8 .. Pesquisas Espaciais, Sao Jose dos Campos, Brazil. . Machado, L. (1998), A fronteira agricola na ~mazoma, m G~ogra­ 'd P M N Leal Jr ., and F. M. Fernandes (1993), Ramfor' . F emusl e, . . , . fia e Meio Ambiente no Brasil, 2nd ed., edited b~ A. Chnstofoest burning and the global carbon budget: Biomass, combustIOn letti et aI., pp 181-217, Hucitec, Sao Paulo, BraZIl. efficiency, and charcoal' formation in the Brazilian Amazon, J. Deforestation in Brazilian Ama· S (2004) , Causes or 'J. d M arguIIS,. Geophys. Res., 98(D9), 16,733-16,743... zon, World Bank, Washington. (Available at http://www-w s. . N C L G Ferreira and F. MlzIara (2007), Deforesta, F errelra, . ., . , worldbank.org). . tion hotspots in the Brazilian Amazon: Evidence and causes as Millikan, B. H. (1992), Tropical deforestation, land degradatIOn, assessed from remote sensing and census data, Earth Interact., and society, lessons from Rondonia, Brazil, Lat. Am. Perspect., 11(1),1-16, doi:l0.1l75/EI201.1. 19(1),45-72. . '1' Hess, L. L., 1. M. Melack, E. M. L. M. N~vo, C. C. F. B.arbosa, oran, E .F, ( 1993) , Deforestation and land use m the Brazllan M and M. Gastil (2003), Dual-season mappmg of wetland mundaAmazon, Human Ecol., 21,1-21. tion and vegetation for the central Amazon basin, Remote Sens. and E. S. Brondizio (1998), Land-use change after M oran, E . F ., . . ki R Environ., 87, 404-428. deforestation in Amazonia, in People and PIxels: Lm ng eSkole C. A. Nobre, J. L. Hackler, K. T. D L H oughton,R"A' " mote Sensing and Social Science, edited by D. Liverman et aI., , Lawrence, and W. H. Chomentowski (2000), Annua~ ~luxes of 94-120 National Academy Press, Washington, D. C. carbon from deforestation and regrowth in the BraZIlian Ama- M~~~n, E. F.: E. Brondiz~o, P. Mausel, and Y. Wu ~1994), In~e­ . zon, Nature, 403, 301-304. grating Amazonian vegetation, land-use and satellite data, BwINPE (2007), Projeto PRODES. (Available at http://www.obt.mpe. Science, 44, 329-338. . br/prodes). Moran, E. F., E. S. Brondizio, J. M. Tucker, M. C. da S~lva-F.o:sberg, Instituto Brasileiro de Geografia e Estatistica (mGE) ~1996), Dados S. McCracken, and I. Falesi (2000), Effects of SOli fertility and do Censo Agropecuario. (Available at http://www.lbge.gov.~r). land-use on forest succession in Amazonia, For. Ecol. Manage., Instituto Nacional de Pesquisas Espaciais (INPE) (2001), MOllltor139,93-108. ing of the Brazilian Amazonian For~s~ by ~atelli~e, Reprint, In- Morton, D. C., R. S. DeFries, Y. E. Shimabukuro, L. O. Anderstituto Nacional de Pesquisas EspacIals, Sao Jose dos Campos, son, F. d. B. Espirito-Santo, M. C. Hansen,. an~ M. Carr~ll Brazil. . (2005) Rapid assessment of annual deforestatIOn m the Brazil. ki E W D C Morton , R. S. DeFries, Y. E. Shimabukuro, ian ~azon using MODIS data, Earth Interact., 9(8), E1139, J asms , . ., . , L. O. Anderson, and M. C. Hansen (2005), Physicalla~dscape doi: 1O.l175/E1139.1. correlates of the expansion of mechanized agricultur~ III Mato Morton, D. C., R. S. DeFries, Y. E. Shimab~kuro, L. O. A~der­ Grosso, Brazil, Earth Interact., 9(16), EI143 , dOl: 10.1175/ son, E. Arai, F. d. B. Espirito-Santo, R. Freitas, ~nd J. Mor~se~e

(2006), Cropland expansion changes deforestatIOn d~namlcs m the southern Brazilian Amazon, Proc. Nat!. Acad. SCI. U. S. A., 103, 14,637-14,641, doi:lO.1073/pnas.0~06377103.. 289-316. '1 Morton, D. C., Y. E. Shimabukuro, R. Freitas, E. Aral, an~ R. S. Kay, K. (2005), Estimating wet-season 'deforestation in the BrazlDeFries (2007a), Secondary forest dynamics and Cerradao ~oss ian Amazon using MOD13 250m data, Geography, M. S., 25, in Mato Grosso during 2001-2005 from MODIS phenology tIm.e Univ. of Maryland, College Park. series, paper presented at XIII Simp. Bras. Sens. ~emoto, Flo~l­ Keller, M., M. A. Silva-Dias, D. C. Nepstad, and M. O..Andre~e anopolis, Sta Catar, Brazil, 21-26 Apr 2007. (Available at http.!/ (2004), The large-scale biosphere-atmosphere expen~ent III www.dsr.inpe.br/sbsr2007Ibibliotecal). . Amazonia: Analyzing regional land use change e~fects, III Eco- Morton, D. c., Y. E. Shimabukuro, B. F. T. Rudorff, A. Lm~a, R. systems and Land Use Change, edited by R. deFnes, G. Asner, Freitas, and R. S. DeFries (2007b), Challenge for conservatIOn at and R. Houghton, pp. 321-334, AGU, Boston, Mass. . the agricultural frontier: Deforestation, fire, and land use dynam· . M A Verissimo , and L. Sobral (2005), Forest Facts m L ent 1m, ., . . . ics in Mato Grosso, Agua Ambiente, 2,5-20. the Brazilian Amazon 2003, Imazon, Belem. . Morton, D. C., R. S. DeFries, J. T. R!\nderson~ L. Glg~lO, Lu, D., P. Mausel, M. Batistella, and E. F. M~ran (2004): ~ompanSchroeder, and G. R. van der Werf (2008), Agrlcultur~l mtenslson of land-cover classification methods III the BraZIlian Amafication increases deforestation fire activity in Amazoma, Global zon Basin, Photogramm. Eng. Remote Sens., 70, 72~-731. . Change Bioi., 14, 2262-2275. Lu, D., M. Batistella, and E. F. Moran (2008), .IntegratlOn ofLa~d­ Morton, D. C., R. S. DeFries, and Y. E. Shimabukuro (2009), Cropsat TM and SPOT HRG images for vegetation change detectIOn land expansion in cerrado and transition forest ~cosystems: in the Brazilian Amazon, Photogramm. Eng. Remote Sens., 73, Quantifying habitat loss from satellite-based vegetation phenol-

EI143.1. . . W (2 006) , Producing a modern agricultural frontier: FIrmS 2 J epson,. and cooperatives in Eastern Mato Grosso, Econ. Geogr., 8 ,

w,.

421-430.

ogy, in Cerrado Land-Use and Conservation: Assessing Tradeoffs Between Human and Ecological Needs, edited by C. Klink, R. S. DeFries, and ~;(;avalcanti, Conservation Int., Washington, D. C., in press. F Nepstad, D. C., et <Jj. (1999), Large-scale impoverishment of Amazonian forestsi{y logging and fire, Nature, 398, 505-508. Numata, I., O. A. Chadwick, D. A. Roberts., 1. P. Schimel, F. F. Sampaio, F. C. ,Leonidas, and 1. V. Soares (2007), Temporal function of soil order, pastIlre age, and management, Rondonia, Brazil,Agric. Ecosyst. Environ., 118, 159-172. Ratana, P., A. R. Huete, and L. G. Ferreira (2005), Analysis of CelTado physiognomies and conversion in the MODIS seasonaltemporal domain, Earth Interact., 9, E1119, doi:l0.1175110873562(2005)0092.0.CO;2. Roberts, D. A., I. Numata, K. Holmes, G. Batista, T. Krug, A. Monteiro, B. Powell, and O. A. Chadwick (2002), Large area mapping ofland-cover change in Rondonia using multitemporal spectral mixture analysis and decision tree classifiers, 1. Geophys. Res., 107(020), 8073, doi:l0.1029/2001JD000374. Sawyer, D. (1984), Frontier expansion and retraction in Brazil, in Frontier Expansion in Amazonia, edited by M. Schmink, and C. H. Wood, pp. 180-203, Univ. of Florida Press, Gainesville. Schmink, M., and C. H. Wood (1992), Contested Frontiers in Amazonia, Columbia Univ. Press, New York. Serrao, E. A. S., and 1. M. Toledo (1990), The search "for sustainability in Amazonian pastures, in Alternatives to Deforestation: Steps Toward Sustainable Use ofthe Amazon Rain Forest, edited by A. B. Anderson, pp. 195-214, Columbia Univ. Press, New York. Shimabukuro, Y. E., V. Duarte, M. A. Moreira, E. Arai, D. M. Valeriano, L. O. Anderson, and F. d. B. Espirito-Santo (2007), Desflorestamento na Amazonia-Sistema DETER, in Sensor MODIS e Suas Aplicar;i5es Ambientais no Brasil, edited by B. F. T. Rudorff, Y. E. Shimabukuro, and J. C. Ceballos, pp. 389-401, Editora Parentese, Sao Jose dos Campos, Brazil. Skole, D., and C. Tucker (1993), Tropical deforestation and habitat fragmentation in the Amazon - Satellite data from 1978 to 1988, Science, 260, 1905-1910. Soares-Filho, B., A. Alencar, D. Nesptad, M. Cerqueira, M. C. V. Diaz, S. Rivero, L. Solorzano, and E. Voll (2004), Simulating the response of land-cover changes to road paving and governance along a major Amazon highway: The Santarem-Cuiaba cOlTidor, Global Change Bioi., 10, 745-764. Souza, C., Jr., L. Firestone, L. M. Silva, and D. Roberts (2003), Mapping forest degradation in the Eastern Amazon from SPOT 4 through spectral mixture models, Remote Sens. Environ., 87, 494-506.

Souza, C., Jr., D. A. Roberts, and M. A. Cochrane (2005), Combining spectral and spatial information to map canopy damages from selective logging and forest fires, Remote Sens. Environ., 98, 329-343. Tardin, A. T., D. C. L. Lee, R. 1. R. Santos, O. R. Assis, M. P. S. Barbosa, M. L. Moreira, M. T. Pereira, D. Silva, and C. P. Santos Filho (1980), Subprojeto Desmatamento: Convenio IBDF/ CNPq-INPE, Technical Report INPE-I649-RPE/I03, Instituto de Pesquisas Espaciais, Sao Jose dos Campos, Brazil. Uhl, C., A. Verissimo, M. M. Mattos, Z. Brandino, and I. C. G. Vieira (1991), Social, economic, and ecological consequences of selective logging in an Amazon frontier-The case of Tailandia, For. Ecol. Manage., 46, 243-273. Veiga, J. B., 1. F. Tourrand, M. G. Piketty, R. Poccard-Chapuis, A. M. Alves, and M. C. Thales (2004), Expansiio e Trajetorias da Pecuaria na Amazonia: Para, Brasil, Editora Universidade de Brasilia, Brasilia, Brazil. Velho, O. G. (1976), Capitalismo Autoritario e Campesinato, DIFEL, Sao Paulo, Brazil. Verissimo, A., P. BalTeto, M. Mattos, R. Tarifa, and C. Uhl (1992), Logging impacts and prospects for sustainable forest management in an old Amazonian frontier-The case of Paragominas, For. Ecol. Manage., 55, 169-199. Warnken, P. F. (1999), The Development and Growth of the Soybean Indusfly in Brazil, Iowa State Univ., Ames. Zarin, D. J., et al. (2005), Legacy of fire slows carbon accumulation in Amazonian forest regrowth, Front. Ecol. Environ., 3, 365-369. I

D. S. Alves, InStltuto Nacional de Pesquisas Espaciais (INPE), DPI (SRE 2), Avenida dos Astronautas 1758, CEP 12227-010, Sao Jose dos Campos SP, Brazil. ([email protected]) M. Batistella, Embrapa Satellite Monitoring, Avenida Soldado Passarinho, 303 CEP 13070-15, Campinas SP, Brazil. (mb@cnpm. embrapa.br) D. C. Morton, Goddard Space Flight Center, 8800 Greenbelt Road, Code 614.4, Greenbelt, MD 20771, USA. (douglas.morton@ nasa.gov, [email protected]) D. A. Roberts, Department of Geography, EH 1832, University of California Santa Barbara, Santa Barbara, CA 93117, USA. ([email protected]) C. Souza Jr., Instituto do Homem e Meio Ambiente da Amazonia (Imazon), Rua Domingos Marreiros 2020, CEP 66060-160, Belem PA, Brazil. ([email protected])

Selective Logging and Its Relation to Deforestation Gregory P. Asner,' Michael Keller,z,3 Marco Lentini,4 Frank Merry,5 and Carlos Souza Jr. 6 Selective logging is a major contributor to the social, economic, and ecological dynamics of Brazilian Amazonia. Logging activities have expanded from lowvolume floodplain harvests in past centuries to high-volume operations today that take about 25 million m 3ofwood from the forest each year. The most common highimpact conventional and often illegal logging practices result in major collateral forest damage, with cascading effects on ecosystem processes. Initial carbon losses and forest recovery rates following timber harvest are tightly linked to initial logging intensity, which drives changes in forest gap fraction, fragmentation, and the light environment. Other ecological processes affected by selective logging include nutrient cycling, hydrological function, and postharvest disturbance such as fire. This chapter synthesizes the ecological impacts of selective logging, in the context of the recent socioeconomic conditions throughout Brazilian Amazonia, as determined from field-based and remote sensing studies' carried out during the Large-Scale Biosphere-Atmosphere Experiment in Amazonia program.

I 1. INTRODUCTION

newab1e resource for the region. There is general consensus that selective logging is widespread and important to the economy; however, the industry has suffered from generally weak and inconsistent government oversight, low capital investment, and a lack of understanding of both forest ecology and management. This combination of conditions has prevented the development of a sustainable logging industry and has led to considerable ecological damage. In the past decade, the ecological, social, and geographic sciences have made important but disparate strides to understand the dynamics of selective logging in Amazonia, with a focus on Brazil where most studies have been conducted. Our goal here is to synthesize the work from these studies and to clarify our understanding of the ecological role of timber production. We focus on the contributions from the Large-scale Biosphere-Atmosphere Experiment in Amazonia (LBA) program. We start with a brief history of the logging industry in Brazil, including the pertinent aspects of the social, economic, and policy drivers oflogging practices. We then link this knowledge of the historical and contemporary conditions for the Amazon forest industry to recent scientific

Selective logging is an important land use in Amazonia. The logging industry is an economic engine that generates revenue, provides jobs, and has the potential to be a re-

'Department of Global Ecology, Carnegie Institution, Stanford, California, USA. 2International Institute of Tropical Forestry, USDA Forest Service, Rio Piedras, Puerto Rico. 3NEON, Inc., Boulder, Colorado, USA. 4Instituto Floresta Tropical, Belem, Brazil. sWoods Hole Research Center, Falmouth, Massachusetts, USA. 6Instituto do Homem e Meio Ambiente da Amazonia, Belem, Brazil.

Amazonia and Global Change Geophysical Monograph Series 186 Copyright 2009 by the American Geophysical Union. 10.1029/2008GM000723

25

26

SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

findings demonstrating the effects of logging on the ecology of the region. Throughout the chapter, we also highlight the contributions of remote sensing as a tool to understand and monitor the course and consequences of selective logging in Amazonia. 2. SOCIOECONOMIC CHARACTERIZATION OF SELECTIVE LOGGING

2.1. Development ofLogging Frontiers in Amazonia European settlers had begun logging the Amazon forest by the seventeenth century [Rankin, 1985]. For the first three centuries of settlement, logging was restricted to low-volume harvest of floodplain forests along the main Amazonian rivers and was of secondary importance to other extractive industries such as Brazil nuts and rubber. It was not until the 1950s that industrial mills, mainly subsidiaries of large international companies such as Georgia Pacific, sprang up in the Amazon estuary to produce high-value sawn wood and veneer for export. Among the earliest examples of selective logging were two floodplain tree species known as virola (Virola surinamensis) and andiroba (Carapa guianensis) [Barros and Uhl, 1995; Pinedo-Vasquez et al., 2001; Zarin et al., 2001]. In the 1960s and 1970s, government policies and investments in infrastructure throughout Amazonia opened access to extensive portions of upland forests [Binswanger, 1991; Browder, 1988; Scholz, 2000]. An extensive and migratory logging industry emerged based on a low-cost raw material in newly forming economic frontiers with minimal governance [Uhl et al., 1997; Verissimo et ai., 1998,2002; Stone, 1998a]. The industry blossomed into a diverse sector with new products and extensive national markets, changing the nature of selective logging along the way. Instead of one or two key species destined for export, a domestic market based on rough-sawn wood and eventually pl~ood absorbed a greater variety of species. Notwithstanding the penetration of mahogany logging deep into the forest, the majority of selective logging operations followed the new roads to harvest high volumes. The new logging strategy created boom and bust economies, severe ecological damage, and a legacy of wasteful and marginally legal practices that still pervade the industry. After 3 decades of deforestation and unplanned selective logging, timber stocks in the old frontiers became largely depleted. Old logging frontiers (Plate 1), which closely follow the arc of deforestation in the states of Para, Mato Grosso, and Rondonia, still encompass 45% of the Amazonian logging centers, but they now generate only about 50% of the revenues and jobs of the timber industry [Lentini et al.,

2005]. The increasing scarcity of raw material stimulated the migration of firms to newer frontiers (intermediate and new frontiers in Plate 1). Roads that strike deep into the interior of Amazonia, mainly the BR-163 Cuiaba-Santarem Highway and, to a lesser extent, the BR-230 Transamazon Highway, have seen a dramatic surge in sawmills and logging. Nepstad et al. [1999] used sawmill surveys conducted by the Brazilian nongovernmental organization Instituto do Homem e Meio Ambiente da Amazonia (IMAZON) and showed that, for the period 1995-1996, logging centers were busy in nearly all states of Brazilian Amazonia (Plate 2a). The pattern of logging centers is similar to the detailed geographic distribution oflogged forest revealed in a remote sensing analysis for the years 1999-2002 (Plate 2b) [Asner et al., 2005]. In 2004, IMAZON catalogued 82 Amazonian logging centers encompassing 3132 timber mills, which consumed 24.5 million m 3 oflogs that produced lOA million m 3 of processed timber including sawn wood, veneers, plywood, and finished wood products (Table 1). This implies an average production yield of only 42% [Lentini et al., 2005]. More than 90% of the production from Brazilian Amazonia is currently concentrated in the states of Para, Mato Grosso and Rondonia. The total gross revenue of the timber industry in Brazilian Amazonia in 2004 was about $2.3 billion U.S. dollars generating approximately 380,000 jobs, including 124,000 direct jobs (processing and logging) and 255,000 indirect jobs. Although there are cost and market share differences between the new, intermediate and old frontiers, this frontier migration has not been accompanied by notable improvements in forest management and timber processing [Merry et al., 2006], as discussed in section 2.2.

2.2. Economics ofSelective Logging In their quest for high-quality raw materials, loggers seek new forest frontiers. The economics of the logging industry in Brazil directly influence the management approaches and therefore strongly affect ecological impacts and the longterm su~tainability of forest timber production. As we mentioned in section 2.1, selective logging has evolved from a single- or few-species model, typical of floodplain logging and mahogany harvest, to a model that can remove up to 40 m 3 ha- 1 and can comprise any variety of 50 or more species. This approach, called conventional logging, is widely used and is profitable. Spatial-economic models estimate the feasible extent of selective logging in Brazilian Amazonia, based on the expected costs for harvest timbe~ and log transport and the prices for logs in the logging centers [Stone, 1998b; Verissimo et al., 1998,2000]. Basically, these models (e.g., Plate 1) identify forests in which selective logging is economically viable and show widespread potential for

ASNER ET AL.

70·VV ~,

65·VV

55· VV

50· VV

45' VV

~

)~7

COLOMBIA

60· VV

27

I

VENEZUELA

~

5· N

"

o

5· S

LoggIng centers o

100,000 to 200,000 m3

•

200,000 to 600,000 m3

•

> 600,000 m3

Logging frontiers

PERU

BOLIVIA

o

Old (> 30 years)

_ _

Intermediary (10-30 years) New « 10 years)

_

Estuarine zone

NMain roads

o

Brazilian Amazon States

Plate 1. Geographic distribution oflogging centers in 2004 and harvestable regi~ns of the Brazilian Amazon [Lentini I et al., 2005].

the forest sector to play an important role in the development of emerging frontiers. Plates 1 and 2 demonstrate that logging centers have expanded sufficiently so that harvestable forests cover most of Brazilian Amazonia. The growth of the logging industry in Brazil has not been determined by new harvesting and processing technologies that are available to logging companies. Instead, the highdamage, high-waste approaches involved in many conventionallogging methods have persisted [Pereira et al., 2002; Holmes et ai., 2002; Asner et al., 2006]. Moreover, despite technological advances in forest management, whereby reduced-impact logging (RIL) methods can be employed to harvest wood while minimizing damage to the forest and maintaining economic returns [Sist, 2000], there has been poor adoption of good management practices. A brief list of reasons why RIL has not been widely adopted includes poorly defined property rights, high transaction costs associated with government bureaucracy, poor distribution of information on good forest management, and an entrenched rentseeking bureaucracy [Putz et al., 2000; Boltz et al., 2001]. It is these influences on firm decision-making that continue to

encourage the use of poor-quality selective logging practices over RIL. As discussed in section 3, this timber-harvesting environment results in ecological responses that have just recently been quantified during the LBA program. While RIL logging has many ecological benefits, the economic benefits of RIL are less certain [Putz et al., 2000]. Among the problems ofRIL logging is, ironically, the preservation of a nearly intact canopy. For all of its biodiversity, microclimatic, and fire protection benefits, an intact canopy keeps potential regenerating trees in the dark and thereby limits postlogging growth. A potential solution to this problem is the selective elimination of competitors around future harvest trees, known as liberation [Wadsworth and Zweede, 2006]. Dauber et al. [2005] modeled tree growth based on field data from an extensive network of plots with growth rates measured on more than 10,000 trees in Bolivian Amazonia. Tree growth was modeled for no treatment and for a silvicultural treatment where surrounding competitive trees and vines are killed for a 25-year cutting cycle in four regions. While modeled first-harvest volumes were considerably larger than the second-harvest volumes for all four

26

SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

findings demonstrating the effects oflogging on the ecology of the region. Throughout the chapter, we also highlight the contributions of remote sensing as a tool to understand and monitor tQe course and consequences of selective logging in Amazonia. 2. SOCIOECONOMIC CHARACTERIZATION OF SELECTIVE LOGGING 2.1. Development ofLogging Frontiers in Amazonia

2005]. The increasing scarcity ofraw material stimulated the migration of firms to newer frontiers (intermediate and new frontiers in Plate 1). Roads that strike deep into the interior of Amazonia, mainly the BR-163 Cuiaba-Santarem Highway and, to a lesser extent, the BR-230 Transamazon Highway, have seen a dramatic surge in sawmills and logging. Nepstad et al. [1999] used sawmill surveys conducted by the Brazilian nongovernmental organization Instituto do Homem e Meio Ambiente da Amazonia (IMAZON) and showed that, for the period 1995-1996, logging centers were busy in nearly all states of Brazilian Amazonia (Plate 2a). The pattern of logging centers is similar to the detailed geographic distribution oflogged forest revealed in a remote sensing analysis for the years 1999-2002 (Plate 2b) [Asner et aI., 2005]. In 2004, IMAZON catalogued 82 Amazonian logging centers encompassing 3132 timber mills, which consumed 24.5 million m3 of logs that produced 10.4 million m3 of processed timber including sawn wood, veneers, plywood, and finished wood products (Table 1). This implies an average production yield of only 42% [Lentini et al., 2005]. More than 90% of the production from Brazilian Amazonia is currently concentrated in the states of Para, Mato Grosso and Rondonia. The total gross revenue ofthe timber industry in Brazilian Amazonia in 2004 was about $2.3 billion U.S. dollars generating approximately 380,000 jobs, including 124,000 direct jobs (processing and logging) and 255,000 indirect jobs. Although there are cost and market share differences between the new, intermediate and old frontiers, this frontier migration has not been accompanied by notable improvements in forest management and timber processing [Merry et al., 2006], as discussed in section 2.2.

European settlers had begun logging the Amazon forest by the seventeenth century [Rankin, 1985]. For the first three centuries ofsettlement, logging was restricted to low-volume harvest of floodplain forests along the main Amazonian rivers and was of secondary importance to other extractive industries such as Brazil nuts and rubber. It was not until the 1950s that industrial mills, mainly subsidiaries of large international companies such as Georgia Pacific, sprang up in the Amazon estuary to produce high-value sawn wood and veneer for export. Among the earliest examples of selective logging were two floodplain tree species known as virola (Virola surinamensis) and andiroba (Carapa guianensis) [Barros and Uhl, 1995; Pinedo-Vasquez et al., 2001; Zarin et al., 2001]. In the 1960s and 1970s, government policies and investments in infrastructure throughout Amazonia opened access to extensive portions of upland forests [Binswanger, 1991; Browder, 1988; Scholz, 2000]. An extensive and migratory logging industry emerged based on a low-cost raw material in newly forming economic frontiers with minimal governance [Uhl et al., 1997; Verissimo et al., 1998,2002; Stone, 1998a]. The industry blossomed into a diverse sector with 2.2. Economics ofSelective Logging new products and extensive national markets, changing the nature of selective logging along the way. Instead of one or In their quest for high-quality raw materials, loggers seek two key species destined for export, a domestic market based new forest frontiers. The economics of the logging industry on rough-sawn wood and eventually pl)"Yood absorbed a in Brazil directly influence the management approaches and greater variety of species. Notwithstanding the penetration therefore strongly affect ecological impacts and the longof mahogany logging deep into the forest, the majority of term sw;tainability of forest timber production. As we menselective logging operations followed the new roads to har- tioned in section 2.1, selective logging has evolved from a vest high volumes. The new logging strategy created boom single- or few-species model, typical of floodplain logging and bust economies, severe ecological damage, and a legacy and mahogany harvest, to a model that can remove up to of wasteful and marginally legal practices that still pervade 40 m3 ha- I and can comprise any variety of 50 or more spethe industry. cies. This approach, called conventional logging, is widely After 3. decades of deforestation and unplanned selective used and is profitable. Spatial-economic models estimate the logging, timber stocks in the old frontiers became largely de- feasible extent of selective logging in Brazili\ln Amazonia, pleted. Old logging frontiers (Plate 1), which closely follow based on the expected costs for harvest timber and log transthe arc of deforestation in the states of Para, Mato Grosso, port and the prices for logs in the logging centers [Stone, and Rondonia, still encompass 45% of the Amazonian log- 1998b; Verissimo et al., 1998,2000]. Basically, these modging centers, but they now generate only about 50% of the els (e.g., Plate 1) identify forests in which selective logging revenues and jobs of the timber industry [Lentini et aI., is economically viable and show widespread potential for

ASNERET AL.

70·W

6S·W

60·W

~,

" ..

OOLOM."~

VENEZUELA

!

SS·W

G-"'·

SO·W

27

4S·W

S· N

N

A o

5· S

LoggIng centers o

100,000 to 200,000 m3

•

200,000 to 600,000 m3

•

> 600,000 m3

Logging frontiers

o

Old (> 30 years)

_

Intermediary (10-30 years)

_

New « 10 years)

_

Estuarine zone

NMainroads

o

Brazilian Amazon States

I

Plate 1. Geographic distribution of logging centers in 2004 and harvestable regipns of the Brazilian Amazon [Lentini et ai., 2005]. I

the forest sector to play an important role in the development of emerging frontiers. Plates 1 and 2 demonstrate that logging centers have expanded sufficiently so that harvestable forests cover most of Brazilian Amazonia. The growth of the logging industry in Brazil has not been determined by new harvesting and processing technologies that are available to logging companies. Instead, the highdamage, high-waste approaches involved in many conventionallogging methods have persisted [Pereira et al., 2002; Holmes et al., 2002; Asner et al., 2006]. Moreover, despite technological advances in forest management, whereby reduced-impact logging (RIL) methods can be employed to harvest wood while minimizing damage to the forest and maintaining economic returns [Sist, 2000], there has been poor adoption of good management practices. A brief list of reasons why RIL has not been widely adopted includes poorl)' defined property rights, high transaction costs associated with government bureaucracy, poor distribution of information on good forest management, and an entrenched rentseeking bureaucracy [Putz et al., 2000; Boltz et al., 2001]. It is these influences on firm decision-making that continue to

encourage the use ofpoor-quality selective logging practices over RIL. As discussed in section 3, this timber-harvesting environment results in ecological responses that have just recently been quantified during the LBA program. While RIL logging has many ecological benefits, the economic benefits of RIL are less certain [Putz et al., 2000]. Among the problems ofRIL logging is, ironically, the preservation of a nearly intact canopy. For all of its biodiversity, microclimatic, and fire protection benefits, an intact canopy keeps potential regenerating trees in the dark and thereby limits postlogging growth. A potential solution to this problem is the selective elimination of competitors around future harvest trees, known as liberation [Wadsworth and Zweede, 2006]. Dauber et al. [2005] modeled tree growth based on field data from an extensive network of plots with growth rates measured on more than 10,000 trees in Bolivian Amazonia. Tree growth was modeled for no treatment and for a silvicultural treatment where surrounding competitive trees and vines are killed for a 25-year cutting cycle in four regions. While modeled first-harvest volumes were considerably larger than the second-harvest volumes for all four

28

SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

ASNER ET AL.

29

Table 1. Socioeconomic Data on Selective Logging in 2004 for Brazilian Amazonia"

,

State Acre Amapa Amazonas Maranhao Mato Grosso Para Rondonia Roraima Combined

SON

O·

S'S

Lo ~ing Centers ( Imber Firms) 1(52) 1 (73) 3 (48) 1 (45) 26 (872) 33 (1,592) 16 (422) 1 (28) 82 (3,132)

Market

Logwood Consumption (x 1000 m 3 a-I)

Gross Income (million U.S. dollars)

Jobs Generated: Direct and Indirect

Exportation (%)

Regional (%)

420 130 490 430 8,010 11,150 3,700 130 24,460

41.6 9.3 55.9 31.7 673.9 1,113.6 368.9 15.9 2,310.7

5,729 2,228 11,344 6,817 108,569 183,741 58,818 2,375 379,621

83 34 64 9 19 50 27 79 36

12 67 18 35 9 11 11 21 11

"Data are retabulated from Lentini et at. [2005].

regions, in the best-case transitional (dry-to-moist forest) ecoregion, the second cut reached 64% of the volume of the first cut under silvicultural treatment compared to only 28% for untreated forest. Silvicultural treatments are costly and currently rarely implemented in Amazonia.

0·8

2.3. Role a/Illegal Logging

N

A b

_

1999-2000 Logging

_

200().2001 logging

_

2001·2002 Logging

_

Federal Conlervatlon Units

_

Indlgenoua ReSeNtS ~

o

__

-~===::11~

500

1.000

~Km

2,000

Plate 2. (a) Regional distribution oflogging centers in the Brazilian Amazon, 1995-1996, derived from sawmill surveys [Nepstad et at., 1999]. Reprinted by permission from Macmillan Publishers Ltd: Nature, copyright 1999. (b) Regional distribution oflogging damage to forests from 1999 to 2002 in the states ofPani (PA), Roraima (RR), Rondonia (RO), Acre (Ae), and northern Mato Grosso (MT) derived from satellite analysis [Asner et at., 2005].

Because it has been widespread, the practice of illegal logging requires some extra attention here. There are two legal mechanisms to gain permission to log forests in Brazilian Amazonia: forest management plans, regulated by specific policy instruments, and deforestation. Current Brazilian law allows the deforestation of 20% of the total area in rural Amazonian properties. In the past, both mechanisms were controlled by the federal environmental agency Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renovaveis (mAMA). Currently, the enactment of a new forestry law (Lei 11284/2006) created the first directives to decentralize the control over forest management plans and to delegate authorizations for deforestation to state agencies in an attempt to increase the transparency of these processes [Zarin et al., 2007]. The few available figures for the log wood volume generated through these mechanisms show that their sum was less than 15 million m 3 between 2000 and 2004. In contrast, surveys of the timber industry carried out in 1998 and 2004 [Lentini et al., 2005] show that the total Amazonian production during this period was above 24 million m 3, a figure directly supported by satellite estimates of Asner et al. [2005]. Combining these estimates, it becomes clear that during these years, at least 40% of the log produc~ tion in Amazonia was illegally generated. In the last several years, government and civil society have taken measures against illegal logging. Hundreds of forest management plans were canceled by mAMA in 2003-2004 in an attempt to halt illegal logging and decrease the defor-

estation rates, estimated by Instituto Nacional de Pesquisas Espaciais (INPE) (the National Space Agency) at 1.7 million ha in 2004 (PRODES: Assessment of deforestation in Brazilian Amazonia, 2005, see http://www.obt.inpe.br/prodes/ index.htrnl). In 2005, the Ministry of Environment, Federal Police, mAMA, and several governmental organizations conducted three large-scale operations targeting illegality in the forest sector and corruption. Despite these recent efforts, the scale of illegal operations remains difficult to precisely I estimate. Illegal logging, conducted without government approved management plan1 and without permits, has especially deleterious economic, social, and ecological effects. Economic impacts can be me.asured by losses in governmental taxes and poor development of economic sectors related to logging. From a societal perspective, formally regulated employment is replaced by informality leading to the low quality ofthese jobs, higher risks for forest workers, lower benefits, and generation of conflicts between loggers and traditional communities over land use. Ecologically, while legal management plans limit logging entries for extended periods (often 30 years), illegal logging promotes multiple logging entries into an area as market conditions change. Multiple entries result in forest impoverishment, a dramatic loss of biodiversity, and increased susceptibility to fire [Nepstad et al., 1999]. 3. ECOLOGICAL IMPACTS OF CURRENT LOGGING PRACTICES The ecological impacts of selective logging are directly related to harvest intensity, in terms of volume of wood removed per hectare and harvest method, which largely determines the level of collateral damage incurred during and after timber harvest. Harvest methods, ranging from largescale conventional logging using crawler tractors and/or wheeled skidders to carefully planned and managed RIL, are

30

SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

tightly linked to both the initial forest damage and to the longer-term ecological responseSTanging from canopy light environment to carbon cycling to fire regimes. The eff&ts of selective logging start with forest biophysical properties, where the forest canopy cover (measured by gap fraction or light availability) is immediately changed during harvest operations. Changes in the spatial pattern and extent of forest gaps then have cascading effects on rates of forest growth, hydrological processes, and the entire food web of a forest. These gaps 'can be arranged by forest-use stratum, including access roads, tree falls where logs are initially cut, skid trails used to remove logs from the forest, and log decks (commonly known as patios) used to load logs onto trucks (Plate 3). These four strata help to organize the ecological impacts as discussed below, and their pattern across the landscape, both in terms of number and density, exerts significant control over the rate of forest recovery for all organisms. Although roads, log decks, skids, and tree fall locations may be visible to satellites for a few years following timber harvest, the canopy rapidly closes (Plate 4), making the geography of logging difficult to track and misguiding the casual observer into thinking that forest recovery is rapid (Appendix A). In fact, while canopy closure is rapid, forest recovery, in terms of both carbon stocks and ecological processes, is far from rapid in upland rain forest areas of Amazonia. Selective logging alters forest ecological processes extending from changes in phenology to the way that carbon and nutrients are cycled and even to the emissions of trace gases such as nitrogen oxides and methane. Although LBA did not focus on tropical forest wildlife, it is important to acknowledge the impacts of logging here because they can be highly variable and often substantial. Several studies have documented how selective logging can cause biotic impoverishment of species and communities or, alternatively, can stimulate population growth of some species [Johns, 1992; Thiollay, 1992; Hill et al., 1995]. Both the magnitude and direction of ecological challge following harvest depend heavily upon the initial logging intensity and the subsequent spatial and temporal dynamics offorest gap fraction (see Fimbel et al. [2001] for an extensive synthesis). 3.1. Light Environment

A major ecological constraint over plant establishment and regeneration in tropical forests is the low light intensity in the understory [Johns et al., 1996]. In humid tropical forests, roughly 2-3% of photosynthetically active radiation (PAR) (400-700 nm) reaches the forest floor [Lee, 1987], and canopy gap fraction typically ranges from just 2 to 4% [e.g., Chazdon and Fetcher, 1984]. Plant productivity or carbon uptake by vegetation is tightly linked to PAR availabil-

ity [Monteith, 1972; Field et al., 1995]. Canopy gaps created by selective timber harvests have immediate effects on PAR interception, latent and sensible heat fluxes, water stress, and plant productivity in tropical forests [Healey et al., 2000; Pinard and Cropper, 2000]. Rates of forest regeneration can thus be linked to the size, number, and spatial arrangement of canopy gaps following harvest [Pereira et al., 2002]. The light environment following logging can be assessed in terms of ground damage resulting from timber harvest (e.g., skids, roads, and decks) and in terms of canopy gap fraction among these types of damage classes. Across a wide range of conventional and RIL harvest intensities (2.6-6.4 felled trees ha- 1) in the eastern Amazon, Asner et al. [2004b] found that the majority of ground damage occurred as skid trails (4-12%), whereas log decks and roads were only a small contributor to the total ground damage «2%). Feldpausch et al. [2005] identified a similar pattern among RIL plots in Mato Grosso, Brazil. However, despite similar harvest intensities, conventional logging causes more canopy damage than does RIL, either from the initial harvest or from tree falls that occur in the years following the logging event [Pereira et al., 2002; Schulze and Zweede, 2006]. Critically, neither the number of log decks nor their individual or total area is well correlated with the number of trees removed or intensity of tree harvesting (trees per hectare) [Asner et al., 2004a]. In contrast, the area of skids is often well correlated with the ground area damaged (square meters) per tree felled, but these features are among the most difficult to map in the field or from satellite sensors (Appendix A). In terms of light interception by the canopy following logging, field surveys across the damage classes show that gap fractions are highest in log decks and lowest in tree fall areas immediately following timber harvest [Feldpausch et al., 2005]. However, the small surface area of log decks reduces their contribution to a very small fraction of the areaintegrated effects of logging throughout a forest. In contrast, lower gap fractions from tree falls are spread throughout a harvested forest, resulting in a large contribution of these areas to the total stand-level canopy gap fraction [Asner et al., 2006]. Canopy openings at tree fall locations are highest at the point where the crown is removed and then decrease with distance from each felled crown (Figure 1). Following harvests in the eastern Amazon, the area affected by the felling of each tree was approximately 100 m in radius for conventional logging but only 50 m for RIL [Asner et al., 2004b]. The size and duration of the tree f~ll gaps have a significant impact on PAR interception and the resulting primary production following timber harvests [Huang et al., 2008]. Pereira et al. [2002] demonstrated the advantages of RIL methods in the maintenance of canopy cover following timber harvest. Feldpausch et al. [2005] observed that RIL

ASNER ET AL.

I[~J

Skids

LZ:J Roads

_

Decf<s • Harvesled Trees

31

I

..'

Plate 3. Plan view of an actual 100 ha logged area in the eastern Brazilian Amazon harvested using conventional methods [Pereira et al., 2002]. r !

reduces canopy damage but only when harvest intensities, calculated in terms of wood volume, are relatively high. In sum, the positive effects of RIL practices on canopy cover are mostly realized at greater harvest volumes; otherwise, they converge on conventional logging damage levels when extraction rates are very low.

the overall carbon laccounting is small. Most of the carbon contained in logs ttansported from forest to mills is rapidly cycled to the atmosphere because sawmill waste is generally burned. Log harvesting operations in the Amazon are inefficient and create a great deal of collateral damage [Verissimo et ~l., 1992]. Table 2 shows that 6 times as much carbon is com3.2. Carbon Cycling mitted to waste in the forest (including fallen and standing coarse woody debris and belowground debris) as is exported The total carbon budget from logging depends upon a bal- as roundwood. In well-managed, low-impact operations, the ance of the long-term storage of carbon in wood products, ratio of debris creation to exported wood can be as low as 2.4 the losses resulting from inefficient milling and processing, [Feldpausch et al., 2005]. The waste includes the portions the losses related to decomposition of wood resulting from of harvested trees that are not trucked out of the forest, trees logging damage to the forest, and the carbon gains over time killed by felling operations, and especially trees killed by from forest regrowth. Carbon lost as a result of the logging the operation of heavy equipment used to open forest roads in the Amazon expressed on an areal basis is summarized in and to skid logs out of the forest. Research indicates that Table 2. Industrial logging in the Amazon region of Brazil skidding operations are responsible for a majority of the colremoves about 19 to 40 m) ha- I of roundwood from the for- lateral damage and that improvements in skidding through est. Compared to other losses, the carbon lost as a result of . the use of proper equipment and planning can cut the damlog processing is small (~12% of the total). Finished prod- age and carbon loss by half [Pereira et al., 2002; Keller et ucts account for only 42% ofthe roundwood removed from al., 2004b]. Most of the waste is generated immediately or the forest [Lentini et al., 2005]. While these products even- within a year of logging. The loss of carbon from the ecotually decay, the lifetime of wood products from Amazon system to the atmosphere is not instantaneous. However, logging is unknown. The relevance of finished products to under the hot and humid conditions of the Amazon, the av-

32

SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

ASNER ET AL.

(b)

(a) OJ

..-

0.5 years postharvest

jO'4 8

b

.

33

I • .RIL 0-

CL

I

~Q. 0.3 &

0.5

r-...--~--'--~----r--======o

1.5 years postharvest 0.4

I•

RIL

--0-

I

CL

0.3

.~0,2

0.2

g ~ 0.1

0.1 0,0

0,0

0

20

40

80

100

80

(c) t:

ts

2.5 years postharvest 0.4

I• --0-

I CL

d

20

40

60

3.5 years postharvest 0.4

~ 0.3

0.3

Cl ~

0.2

80

100

----,,...--====:-1

0,6 ,-.,.....--.--.......

RIL

f!

c

o

(d) 0.6

.2

'---'-------'--~~--'------'--~'----'

I•

RIL

--0-

I

CL

eu

0.2

0

lij

U 0.1 0.0 20

0

40

80

100

80

Distance from felled tree crown (m)

o

20

40

60

80

100

Distance ffm felled tree crown (m)

Figure 1. Mean (plus/minus standard deviation) forest canopy gap fraction at incr~asing distance from felled tree crowns in conventionally logged (CL) and reduced-impact logging (RIL) sites in the eastern Amazon at (a) 0.5, (b) 1.5, (c) 2.5, and (d) 3.5 years following timber harvest [Asner et al., 2004b].

e

f

erage lifetime of coarse woody debris is about 5 to 7 years [Chambers et al., 2000; Palace et al., 2007]. Despite this basic understanding of carbon losses following timber harvest, the primary sources of data remain limited, and additional measurements are needed.

The belowground contribution to the carbon balance is especially uncertain because of the precarious knowledge of belowground carbon stocks in the Amazon [Keller et ai., 2001]. Root stocks are poorly quantified, and it is difficult to quantify small changes in soil carbon pools. In one study,

Table 2. Estimates of Carbon Loss From Logginga Mean (Mg C ha- 1)

Plate 4. Sequence ofLandsat images showing roads and log decks generated by selective logging in central Mato Grosso and the rapid apparent recovery of the canopy, even following fire [Souza et al., 2005]. Colors indicate different normalized difference fraction index from high canopy cover in greens to low canopy cover in pinks and whites. ,

II

... !

Roundwood Aboveground woody debris Belowground woody debris . Standing dead Total

5

26 6 5 42

Low (Mg C ha- ' )

High (Mg C ha- ' )

4 21 5 4 33

8 32 8 6 55

aRoundwood estimates for mean, high, and low harvest volumes are from Asner et al. [2005] and Nepstad et al. [1999]. Roundwood density is assumed to be 0.7 Mg m3, and the conversion factor for wood mass to C is 0.5 [Schlesinger, 1997]. The roundwood loss is adjusted to account for 42% of the wood t4at becomes durable products [Lentini et al., 2005]. Coarse woody debris estimates are based on the work of Keller et al. [2004b]. Belowground loss is calculated as 20% of aboveground coarse woody debris plus roundwood [Keller et al., 2001]. Standing dead is 20% ofaboveground coarse woody debris [Palace et al., 2007].

ASNER ET AL.

34 SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

.:. .l.

researchers quantified soil C02 flux following reducedimpact logging in the Tapaj6s National Forest of central Para State [Keller et al., 2005]. The authors found a slight increase in ~02 emissions from tree fall gaps and skid trails and a decrease in emissions from log decks, compared to background forest emissions. When the C02 emissions were aggregated spatially, the emissions from logged forest were statistically indistinguishable from background emissions. It is likely that increased C02 production related to the decomposition of dead roots is offset 'by the loss of root respiration resulting in no net change in C02 flux [Varner et ai., 2003; Silver et al., 2005]. The rate of regrowth in logged forests in the Amazon has been measured in only a few very limited cases in experimental plots, and those regrowth estimates have focused on estimation of future timber production [e.g., Silva et al., 1995]. We are not aware of any published studies of regrowth rates and carbon budgets following conventional industrial logging in the Amazon. Regrowth rates depend on a variety offactors dominated by light availability (gap fraction) and the efficiency oflight use by the canopies [Huang et ai., 2008]. In general, carbon uptake rates will be highest where a greater portion of the canopy has been opened. Therefore, gross oarbon uptake will be highest where the largest amount of slash has been generated, and this is where the largest gross carbon losses will be incurred from decomposition. Optimistic modeling scenarios suggest that timber

production may be maintained for up to 200 years under 30year rotations, provided that the market accepts new potentially commercial species with time [Alder and Silva, 2000; Keller et ai., 2004a]. However, without substantial silvicultural intervention [Dauber et al., 2005], fast growing species with low wood density will tend to replace slow growing species with high wood density, leading to forest stands with lower total carbon stocks [Keller et al., 2004a; Bunker et al., 2005]. In their simulations, Keller et al. [2004a] and Huang et al. [2008] predicted that Amazon forests would lose an average of 12-19 Mg C ha- I over the first 30 years of rotation (a new area is cut each year) and between 16 and 30 Mg C ha- I with reentry logging cycles over 200 years (Figure 2). It is important to point out that these models are based on average growth rates. Brienen and Zuidema [2007] have shown that skewed distributions ofthe tree growth should be considered for more realistic simulations of tropical forest production. 3.3. Other Biogeochemical Cycles

Beyond its direct effects on forest structure and the carbon cycle, selective logging can alter nutrient cycles and other key biogeochemical processes regulating forest productivity and neighboring aquatic systems. For example, nitrate losses from logged areas in Guyana vary in proportion to the area of soil disturbance surrounding harvested trees [Brou-

200 180 160 10

..c 140

U

0)

e.

120

1 year) very challenging [Stone and Lefebvre, 1998]. During LBA, remote sensing studies on logging in the Brazilian Amazon found that Landsat reflectance data have limited spectral resolution to detect logged forest from intact forest [Asner et al., 2002a; Souza et al., 2005]. Vegetation indices [Stone and Lefebvre, 1998; Souza et al., 2005; Broadbent et al., 2006] and texture filters [Asner et al., 2002a] also showed a limited capability for detection of logging. Improving the spatial resolution of reflectance data can help; Ito 4-m resolution IKONOS satellite data can readily detect forest canopy structure and canopy damage caused by selective logging [Asner et al., 2002b; Read et al., 2003; Souza et al., 2003]. However, the high cost of these images and additional computational challenges of extracting information severely limit the operational use of IKONOS and similar imagery. LBA research showed that the detection of logging at moderate spatial resolution is best accomplished at the subpixel scale using a technique called spectral mixture analysis

Table.At. Remote Sensing Techniques Applied to Selective Logging in the Br '1' A " Mappmg aZI Ian mazon

I:

Approach Studies yo I Isua Watrin and Rocha [1992J interpretation

I,

J:

.

~.

1 :1l J

Betection of logging landings and buffer

Decision tree

Change detection

CLASlite

NDFIand CCA

n

I L.

...I

r

orma Ize

Sensor Landsat TM

Stone and Lefebvre [1998J

LandsatTM

Matricardi et al. [2001J

LandsatTM

Santos et al. [2001J

LandsatTM

I ocal

Souza et al. [2003J

Souza et al. [2003J

SPOT-4

LandsatTM andETM+

LandsatTM

local

local

local

LandsatTM andETM+

http://claslite.ciw.edu

LandsatTM anYWhere andETM+, in the SPOT-415, world ASTER, ALI, and MODIS LandsatTM andETM+

map total logging area

Brazilian Amazon Brazilian Amazon local map total logging area (canopy dam age, clearings, and undamaged forest)

Asner et al. [2005, 2006J

Souza et al. [2005J

Objective

local

Souza and Barreto [2000J, LandsatTM Matricardi et al. [2001 J, andETM+ and Moilteiro et al. [2003J

Grar;a et al. [2005J Image segmentation

CLAS

Spatial Extent

five states of the Brazilian Amazon

local

map forest canopy damage associated with logging and burning map forest canopy damage associated with logging and burning

ASNER ET AL. i

Advantages

Disadvantages

It does not require sophisticated image processing techniques.

It is relatively simple to implement and satisfactory to estimate the total potential logging area. It has simple and intuitive classification rules.

It enhances forest canopy damaged areas.

It is laborintensive for large areas and may be user-biased to define the boundaries of logged forest.

Logging buffers vary across the landscape and do not reproduce the actual shape of the logged area. It has not been tested in very large areas, and classification rules may vary across the landscape. It requires two pairs of images and does not separate natural and anthropogenic forest changes. It has not been tested in very large areas, and segmentation rules may vary across the landscape.

map total It is relatively logging area simple to (canopy damage, implement and clearings, and satisfactorily undamaged estimate the total forest) logging area. Free software is available. It is highly It requires high automated and computation standardized to power and pairs very large areas. of images to detect forest change. It is highly It is limited to automated, run tropical forests. on a standard desktop computer, requires minimal training. It enhances forest canopy damaged areas.

39

It has not been tested in very large areas and does not separate

d d'ffi ' ,ance ematlc Ma PI C I erence fraction index; and CCA Contextual Clas'fi t' APpe~ us; LAS, Carnegie Landsat Analysis System' NDFI , Sl ca IOn Igonthm. ' ,

(SMA). Images obtained with SMA that show detailed fractional cover of soils, nbnphotosynthetic vegetation (NPV), and green vegetatiorr'tmhance our ability to detect logging infrastructure andjanopy damage. For example, soil fractional cover map~erived from SMA can enhance the detection oflog decks and roads [Souza and Barreto, 2000]; maps of NPV fraction enhance the expression of damaged and dead vegetation [Souza et al., 2003]; and the green vegetation fractional cover is sensitive to canopy openings [Asner et al., 2004a]. Several mapping techniques were tested and applied in local to large regional-scale studies of selective logging in Brazil (Table AI). These techniques vary in terms of the mapping goals, the approach and geographic extent, and reported accuracies. In terms of mapping goals, some techniques were developed to map the total potentially logged area, which includes forest canopy damaged and forest clearings and undamaged forest islands, while others were focused only on the mapping of forest canopy damage. The former group of techniques included visual interpretation of Landsat images [e.g., Stone and Lefebvre, 1998; Matricardi et at., 2001], manual and automated detections of)og decks with an estimated timber-harvesting buffer around the decks [Souza and Barreto, 2000], and highly automated SMA approaches combined with pattern recognition methods [Souza et at., 2005; Asner et at., 2005, 2006]. Future studies will likely focus on techniques that balance issues of satellite image quality, availability and cost, processing time, and the level of expertise required to produce verifiable maps of selective logging. LBA research paved the way for these current and future developments. AclO1owledgments. We thank the many individuals and agencies from Brazil, the United States, and elsewhere for years of financial, logistical, and scientific support required to develop an understanding of land use change and logging practices in the Amazon region. This work was supported by the NASA LBA-ECO program and the Gordon and Betty Moore Foundation.

REFERENCES Alder, D., and 1. N. M. Silva (2000), An empirical cohort model for management of terra fume forests in the Brazilian Amazon, For. Ecol. Manage., 130,141-157. Asner; G. P., M. Keller, R. Pereira, and 1. Zweede (2002a), Remote sensing of selective logging in Amazonia: Assessing limitations based on detailed field measurements, Landsat ETM+ and textural analysis, Remote Sens. Environ., 80, 483--496. Asner, G. P., M. Palace, M. Keller, R. Pereira Jr., J. N. M. Silva, and J. C. Zweede (2002b), Estimating canopy structure in an Amazon forest from laser rangefinder and IKONOS satellite observations, Biotropica, 34, 483--492.

Asner, G. P., M. Keller, R. Pereira, 1. C. Zweede, and 1. N. M. Silva (2004a), Canopy damage and recovery following selective logging in an Amazon forest: Integrating field and satellite studies, Ecol. Appl., 14,280-298. Asner, G. P., M. Keller, andJ. N. M. Silva (2004b), Spatial and temporal dynamics of forest canopy gaps following selective logging in the eastern Amazon, Global Change Bioi., 10, 765-783. Asner, G. P., D. E. Knapp, E. N. Broadbent, P. J. C. Oliveira, M. Keller, and 1. N. M. Silva (2005), Selective logging in the Brazilian Amazon, Science, 310,480--482. Asner, G. P., E. N. Broadbent, P. J. C. Oliveira, D. E. Knapp, M. Keller, and J. N. Silva (2006), Condition and fate oflogged forests in the Brazilian Amazon, Proc. Natl. Acad. Sci. U S. A., 103, 12,947-12,950. Barreto, P., C. Souza Jr., R. Noguer6n, A. Anderson, and R. Salomiio (2006), Human Pressure on the Brazilian Amazon Forests, 84 pp., World Resour. Inst., Washington, D. C. Barros, A. C., and C. Uhl (1995), Logging along the Amazon River and estuary: Patterns, problems and potential, For. Ecol. Manage., 77,87-105. Binswanger, H. P. (1991), Brazilian policies that encourage deforestation in the Amazon, World Dev., 19, 821-829. Boltz, F., D. R. Carter, T. P. Holmes, and R. Pereira Jr. (2001), Financial returns under uncertainty for conventional and reducedimpact logging in permanent production forests of the Brazilian Amazon, Ecol. Econ., 39, 387-398. Brienen, R. J. W., and P. A. Zuidema (2007), Incorporating persistent tree growth differences increases estimates of tropical timber yield, Front. Ecol. iEnviron., 5, 302-306. Broadbent, E. N., D:. 1. Zarin, G. P. Asner, M. Pena-Claros, A. Cooper, and R. Littell (2006), Forest structure and spectral properties after selectiVe logging in Bolivia, Ecol. Appl., 16, 11481163. Broadbent, E. N., G. P. Asner, M. Keller, D. E. Knapp, P. J. C. Oliveira, and J. N. Silva (2008), Forest fragmentation and edge effects from deforestation and selective logging in the Brazilian Amazon, Bioi. Conserv., 141, 1745-1757, doi:l0.10161 j .biocon.2008 .04.024. . Brouwer, L. C. (1996), Nutrient Cycling in Pristine and Logged Tropical Rain Forest, Tropenbos-Guyana Ser., vol. 1, TropenbosGuyana Programme, Georgetown, Guyana. Browder, J. O. (1988), Public policy and deforestation in the Brazilian Amazon, in Public Policies and the Misuse of the Forest Resource, edited by R. Repetto and M. Gillis, pp. 247-298, Cambridge Univ. Press, Cambridge, U. K. Bunker, D. E., F. DeClerck, 1. C. Bradford, R. K. Colwell, 1. Perfecto, O. L. Phillips, M. Sankaran, and S. Naeem (2005), Species loss and aboveground carbon storage in a tropical forest, Science, 310,1029-1031. Bustamante, M. M. C., M. Keller, andD. A. da Silva (2009), Sources and sinks of trace gases in Amazonia and the cerrado, Geophys. Monogr. Ser., doi: 1O.l029/2008GM000733, this volume. Cardille, J. A., and J. A. Foley (2003), Agricultural land-use change in Brazilian Amazonia between 1980 and 1995: Evidence from integrated satellite and census data, Remote Sens. Environ., 87, 551-562.

40

SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

ASNERET AL.

41

I

... I

..1 .~

,

"

J. '

Chambers, J. Q., N. Higuchi, J. P. Schimel, L. V. Ferreira, and J. M. Melack (2000), Decomposition and carbon cycling of dead trees in tropical forests of the central Amazon, Oecologia, 122, 380-388..,. Chambers, J. Q., G. P. Asner, D. C. Morton, L. O. Anderson, S. S. Saatchi, F. D. B. Espirito-Santo, M. Palace, and C. Souza Jr. (2007), Regional ecosystem structure and function: Ecological insights from remote sensing of tropical forests, Trends Ecol. Evol.:22,414-423. Chazdon, R. L., and N. Fetcher ~1984), Photosynthetic light environments in a lowland tropical rain forest in Costa Rica, J Ecol., 72,553-564. Cochrane, M. A, A Alencar, M. D. Schulze, C. M. Souza, D. C. Nepstad, P. Lefebvre, and E. A Davidson (1999), Positive feedbacks in the fire dynamic of closed canopy tropical forests, Science,284,1832-1835. Dauber, E., T. S. Fredericksen, and M. Pena-Claros (2005), Sustainability of timber harvesting in Bolivian tropical forests, For. Ecol. Manage., 214, 294-304. Fearnside, P. M., and W. M. Guimaraes (1996), Carbon uptake by secondary forests in Brazilian Amazonia, For. Ecol. Manage., 80,35-46. Feldpausch, T. R, S. Jirka, C. A M. Passos, F. Jasper, and S. Riha (2005), When big trees fall: Damage and carbon export by reduced impact logging in southern Amazonia, For. Ecol. Manage., 219,199-215. Field, C. B., J. T. Randerson, and C. M. Malmstrom (1995), Global net primary production: Combining ecology and remote sensing, Remote Sens. Environ., 51, 74-88. Fimbel, R A., A Grajal, and J. G. Robinson (Eds.) (2001), The Cutting Edge: Conserving Wildlife in Logged Tropical Forests, 700 pp., Columbia Univ. Press, New York. Gerwing, J. (2002), Degradation offorests through logging and fire in the eastern Brazilian Amazon, For. Ecol. Manage., 157, 131-141. Gra9a, P. M., L. A Santos, J. R. Soares, and J. V. Souza (2005), Desenvolvimento metodol6gico para detec91io e mapeamento de areas florestais sob explora9ao madeireira: Estudo de caso, regiao norte do Mato Grosso, in XII Simposio Brasileiro de Sensoriamento Remoto, pp. 1555-1562, Inst. Nac. Pesqui. de Espaciais, Sao Jose dos Campos, Brazil. Healey, J. R, C. Price, and J. Tay (2000), The 'cost of carbon retention by reduced impact logging, For. Ecol. Manage., 139, 237-255. Hill, J. K, K C. Hamer, L. A. Lace, and W. M. T. Banham (1995), Effects of selective logging on tropical forest butterflies on Buru, Indonesia, J App!. Eeal., 32, 754-760. Holmes, T. P., G. M. Blate, J. C. Zweede, R Pereira Jr., P. B. Barreto, F. Boltz, and R. Bauch (2002), Financial and ecological indicators of reduced impact logging performance in the eastern Amazon, For. Ecol. Manage., 163, 93-110. Huang, M., G. P. Asner, M. Keller, and J. A Berry (2008), An ecosystem modelfor tropical forest disturbance and selective logging, J Geophys. Res., 113, G01002, doi:l0.l029/2007JG000438. Johns, AD. (1992), Vertebrate responses to selective logging: Implications for the design of logging systems, Philos. Trans. R. Soc. London, Ser. B, 335,437-442.

Johns, J. S., P. Barreto, and C. Ubi (1996), Logging damage during planned and unplanned logging operations in the eastern Amazon, For. Eco!. Manage., 89, 59-77. Keller, M., and W. A Reiners (1994), Soil-atmosphere exchange of nitrous oxide, nitric oxide, and methane under secondary succession of pasture to forest in the Atlantic lowlands of Costa Rica, Global Biogeochem. Cycles, 8, 399-409. Keller, M., M. Palace, and G. Hurtt (2001), Biomass estimation in the Tapajos National Forest, Brazil-Examination of sampling and allometric uncertainties, For. Ecol. Manage., 154, 371-382. Keller, M., G. P. Asner, J. N. M. Silva, and M. Palace (2004a), Sustainability of selective logging of upland forests in the Brazilian Amazon: Carbon budgets and remote sensing as tools for evaluation oflogging effects, in Working Forests in the Tropics: Conservation through Sustainable Management?, edited by D. Zarin et aI., pp. 41-63, Columbia Univ. Press, New York. Keller, M., M. Palace, G. P. Asner, R. Pereira, and J. N. M. Silva (2004b), Coarse woody debris in undisturbed and logged forests in the eastern Brazilian Amazon, Global Change BioI., 10, 784-795. Keller, M., R K Varner, J. Dias, H. Silva, P. Crill, R de Oliveira Jr., and G. P. Asner (2005), Soil-atmosphere exchange of nitrous oxide, nitric oxide, methane, and carbon dioxide in logged and undisturbed forest in the Tapajos National Forest, Brazil, Earth Interact., 9(23), E1I25, doi: 10. 1175/E1I25. 1. Lee, D. W. (1987), The spectral distribution of radiation in two neotropical rainforests, Biotropica, 19, 161-166. Lentini, M., D. Pereira, D. Celentano, and R. Pereira (2005), Fatos Florestais da Amazonia, Inst. do Homem e Meio Ambiente da Amazonia, Belem, Brazil. Matricardi, E. AT., D. L. Skole, M. A Chomentowski, and M. A Cochrane (2001), Multi-temporal detection of selective logging in the Amazon using remote sensing, Spec. Rep. BSRSI Res. Adv. RA03-01\w, 27 pp., Trop. For. Inf. Cent., Mich. State Univ., East Lansing. McNabb, K L., M. S. Miller, B. G. Lockaby, B. J. Stokes, R G. Clawson, J. A Stanturf, and J. N. M. Silva (1997), Selection harvests in Amazonian rainforests: Long-term impacts on soil properties, For. Ecol. Manage., 93,153-160. Merry, F., G. Amacher, D. Nepstad, P. Lefebvre, E. Lima, and S. Bauch (2006),Industrial development on logging frontiers in the Brazilian Amazon, Int. J Sustain. Dev., 9, 277-296. Monteiro; A L., C. M. Souza Jr., and P. Barreto (2003), Detection of logging in Amazonian transition forests using spectral mixture models, Int. J Remote Sens., 24,151-159. Monteith, J. L. (1972), Solar radiation and productivity in tropical ecosystems, J Appl. Ecol., 9, 747~766. Nepstad, D. C., et al. (1999), Large-scale impoverishment of Amazonian forests by logging and fire, Nature, 398, 505-508. Olander, L. 0., M. M. Bustamante, G. P. Asner, E. Telles, Z. Prado, and B. P. Camargo (2005), Surface soil changes following selective logging in an eastern Amazon forest, Earth Interact., 9(4), E1I35, doi:l0.l175/E1I35.1. Palace, M., M. Keller, G. P. Asner, J. N. M. Silva, and C. Passos (2007), Necromass in undisturbed and logged forests in the Brazilian Amazon, For. Ecol. Manage., 238, 309-318.

Pereira, R., Jr., J. Zweede, G. P. Asner, and M. Keller (2002), Forest canopy damage and recovery in reduced-impact and conventional selective loggiflg in eastern Para, Brazil, For. Eco!. Manage., 168, 77-89.Y Peres, C. A. (2001)/ Synergistic effects of subsistence hunting and habitat fragmed'~tion oil Amazonian forest vertebrates, Conserv. BioI., 15,1490-1505. Peres, C. A, J. Barlow, and W. F. Laurance (2006), Detecting anthropogenic disturbance in tropical forest, Trends Ecol. Evo!., 21,227-229. Pinard, M. A, and W. P. Cropper (2000), Simulated effects oflogging on carbon storage in dipterocarp forest, J App!. Ecol., 37, 267-283. Pinedo-Vasquez, M., D. J. Zarin, K Coffey, C. Padoch, and F. Rabelo (2001), Post-boom logging in Amazonia, Hum. BioI., 29, 219-239. Putz, F. E., D. P. Dykstra, and R. Heinrich (2000), Why poor logging practices persist in the tropics, Conserv. BioI., 14, 951956. Rankin, J. M. (1985), Forestry in the Brazilian Amazon, in Amazonia, edited by G. Prance and T. Lovejoy, pp. 369-392, Pergamon, Oxford, U. K Ray, D., D. Nepstad, and P. Moutinho (2005), Micrometeorological and canopy controls of fire susceptibility in a fOl:ested Amazon landscape, Ecol. Appl., 15,1664-1678. Read, J. M., D. B. Clark, E. M. Venticinque, and M. P. Moreira (2003), Application of merged I-m and 4-m resolution satellite data to research and management in tropical forests, J Appl. Ecol., 40, 592-600. Sanchez, P. A (1976), Properties and Management ofSoils in the Tropics, 235 pp., John Wiley, New York. Santos, J. R., T. Krug, L. S. Araujo, L. G. Meira Filho, and C. A Almeida (2001), Dados multitemporais TMlLandsat aplicados ao estudo da dinamica de explora91io madeireira na Amazonia, inX Simposio Brasileiro de Sensoriamento Remoto, pp. 17511755, Inst. Nac. Pesqui. de Espaciais, Sao Jose dos Campos, Brazil. Schlesinger, W. H. (1997), Biogeochemistry: An Analysis of Global Change, 2nd ed., 588 pp., Academic, San Diego, Calif. Schneider, R, E. Arima, A Verissimo, P. Barreto, and C. Souza Jr. (2000), Sustainable Amazon: Limitations and Opportunities for Rural Development, Inst. do Homem e Meio Ambiente da Amazonia, Brasilia. Scholz, I. (2000), Overexploitation or Sustainable Management: Action Patterns of the Tropical Timber Industly: The Case of Para, Brazil, 1960-1997,441 pp., Frank Cass, London. Schulze, M., and J. Zweede (2006), Canopy dynamics in unlogged and logged forest stands in the eastern Amazon, For. Ecol. Manage., 236, 56-64. Silva, J. N. M., J. O. P. de Carvalho, J. Lopes, B. F. de Almeida, D. H. M. Costa, L. C. de Oliveira, J. K Vanclay, and J. P. Skovs~ gaard (1995), Growth and yield of a tropical rain forest in the Brazilian Amazon 13 years after logging, For. Ecol. Manage., 71,267-274. Silver, W. L., J. Neff, M. McGroddy, E. Veldkamp, M. Keller, and R Cosme (2000), Effects of soil texture on belowground carbon

and nutrient storage in a lowland Amazonian forest ecosystem, Ecosystems, 3,193-209. Silver, W. L., A W. Thompson, M. E. McGroddy, R. K. Varner, J. D. Dias, H. Silva, P. M. Crill, and M. Keller (2005), Fine root dynamics and trace gas fluxes in two lowland tropical forest soils, Global Change BioI., 11, 290-306. Sist, P. (2000), Reduced-impact logging in the tropics: Objectives, principles and impacts, Int. For. Rev., 2, 255-263 . Souza, C., and P. Barreto (2000), An alternative approach for detecting and monitoring selectively logged forests in the Amazon, Int. J Remote Sens., 21,173-179. Souza, C., L. A Firestone, L. Moreira, and D. A Roberts (2003), Mapping forest degradation in the eastern Amazon from SPOT 4 through spectral mixture models, Remote Sens. Environ., 87, 494-506. Souza, C. M., D. A Roberts, and M. A. Cochrane (2005), Combining spectral and spatial information to map canopy damage from selective logging and forest fires, Remote Sens. Environ., 98, 329-343. Steininger, M. K (1996), Tropical secondary forest regrowth in the Amazon: Age, area, and change estimation with thematic mapper data, Int. J Remote Sens., 17, 9-27. Stewart, J. W. B., and H. Tiessen (1987), Dynamics of soil organic phosphorus, Biogeochemistry, 4, 41-60. Stone, S. W. (l998a), Evolution of the timber industry along an aging frontier: The case of Paragominas (1990-95), World Dev., 26,433-448. Stone, S. W. (1998b), Using a geographic information system for applied policy analysis: The case of logging in the eastern Amazon, Ecol. Econ., p7, 43-61. Stone, T. A, and P. Lefebvre (1998), Using multi-temporal satellite data to evaluate selective logging in Para, Brazil, Int. J Remote Sens., 19,2517-2526. Thiollay, J. M. (1992), Influence of selective logging on bird species diversity in a Guiana rain forest, Conserv. Bioi., 6, 47-63. Ubi, C., P. Barreto, A Verissimo, E. Vidal, P. Amaral, A C. Barros, C. Souza, J. Johns, and J. Geiwing (1997), Natural resource management in the Brazilian Amazon, BioScience, 47,160-168. Varner, R K, M. Keller, J. R Robertson, J. D. Dias, H. Silva, P. M. Crill, M. McGroddy, and W. L. Silver (2003), Experimentally induced root mortality increased nitrous oxide emission from tropical forest soils, Geophys. Res. Lett., 30(3), 1144, doi: 10.1 029/2002GLO16164. Verissirno, A., P. Barreto, M. Mattos, R. Tarifa, and C. Ubi (1992), Logging impacts and prospects for sustainable forest management in an old Amazonian frontier: The case of Paragominas, For. Ecol. Manage., 55,169-199. Verissimo, A, C. S. Junior, S. Stone, and C. Ubi (1998), Zoning of timber extraction in the Brazilian Amazon, Conserv. Bioi., 12,128-136. Verissimo, A., C. Souza, and P. Amaral (2000), Identifica9ao de .Areas com Potencial para a Cria9ao de Florestas Nacionais na Amazonia Legal, report, 36 pp., Braz. Minist. of Environ., Brasilia. (Available at http://www.imazon.org.br/upload/im_ livros_OIO.zip)

42

I,

.. .Ii

SELECTIVE LOGGING AND ITS RELATION TO DEFORESTATION

Verissimo, A., E. Lima, and M. Lentini (2002), Palos Madeireiros do Estado do Pani, report, 72 pp., Inst. do Homem e Meio Ambiente da Amazonia, Belem, Brazil. (Available at http://www. imazon.oJg.br/publicacoes/publicacao.asp?id=Ill) Vitousek, P. M., and R. L. Sanford Jr. (1986), Nutrient cycling in moist tropical forest, Annu. Rev. Eco!. Syst., 17, 137-167. Wadsworth, F. H., and J. C. Zweede (2006), Liberation: Acceptable production of tropical forest timber, For. Ecol. Manage., 233,45-51. Watrin, O. S., and A. M. A. Rocha (1992), Levantamento da vegeta

.lll 40 c Q)

e

~

80000 60000

2002

2003

2004

_VCFO-20

_VCF20-40

c::=:::::J VCF 60-80

-

2005 _VCF40-60

-HotSpots

lK

Annual Cloud Fraction

2000

45000 ,1;;1

1800

40000

,E!'

1600

35000

~ ~ 1400

C 0

19

~

-5

1::

E

'iii

,

30000 ~

'"

.!!!- 1200

~

.l!l

.c 0

1000

z

25000

tI

'0

O'

'0 o

C

'§

~

600

8.

(f)

20000 ~

0>

800

15000

Qi

E ::J

::J

Z

~

o 2001

'0

a.

20000

0

::c '0 Qi ..c

-oC/)

40000

10

tli

~ (f)

30 20

~

53

400

10000

200

5000

o

Z

o 1998 _

1999 2000

2001

2002

2003

2004

2005

2006

Roraima _Acre· . 0 .. Brazilian Amazonia

10'8

Figu~e 1. (a) MOD~S0erra daytime and nighttime fire product summary statistics during 2001-2005. (dashed line) Hot spot mter~nnual v~natlOn pres~nted as totals and (vertical bars) as annual percentage distribution using unifonn Vegeta-

o 125250

500

tIOn Contmuous FIelds (VCF) mtervals. (asterisks) Basin-wide average cloud-cover fraction derived from MODISlTerra shows little interannual variation «4%). (b) (dashed line) Annual number of hot spots detected by the TRMMNIRS instmment over the Brazilian Amazonia. Vertical bars describe hot spots detected over Roraima and Acre states, which were severely affected by the 1997-1998 EI Nino and the 2005 wanning ofthe tropical North Atlantic waters, respectively.

750

IKilome(ers

i 70'W

60'W

50'W

Plate 2. Map of peak month of hot spot detection as observed by MODISlTerra and Aqua, AVHRR-NOAA 12 and GOES 08, and 12 daytime and nighttime images during 2003-2006 (4.15 million hot spots processed). Data was aggregated using a 0.05° x 0.05° grid. Only cells having hot spot detections produced from two or more sensors (medium and high confidence cells from Plate la) are represented on the map.

which may develop in parts of the region. Increased forest flammability associated with higher risk of fires escaping control often lead to widespread forest fires affecting significantly large areas [Van der Werfet al., 2004; Alencar et aI.,

2006; Brown et al., 2006; Nepstad et al., 1999a]. Satellite active fire products will normally show strong peaks departing from the annual average in fire activity associated with such large-scale climate anomalies (Figure 1).

-----1'

1

II

j

54

...

:ri... .Ii

.Ii !!

...

SPATIAL DISTRIBUTION AND INTERANNUAL VARIABILITY OF FIRE IN AMAZONIA

Some of the most recent alternatives designed to cope with the increase in fire activity in Brazilian Amazonia were associated with the adoption of specific public policies including tempm-ary fire prohibition, increase in law enforcement, and the creation of new conservation areas. Fire prohibition was first proposed for Mato Grosso state in 2001 and was followed by other regions in subsequent years. It is used as a preventive mechanism limiting burning at the peak of the dry season as well as an emergency response when rapid reduction in fire activity is desired [Brown et al., 2006]. Successful application of a fire moratorium depends on the effectiveness oflaw enforcement and on community engagement. The creation of new conservation areas also depends on the effectiveness of law enforcement and park administration, and in some cases, the established areas may not withstand the threats of logging and fire [Ferreira et al., 1999; Laurance and Williamson, 2001; Pedlowsld et al., 2005]. Pressure is building along conservation units where the surrounding forests are being depleted (see examples in Plate 1). 6. SPATIAL AND NUMERICAL RELATIONSHIPS WITH DEFORESTATION RATES As described above, vegetation fires and deforestation activities in Amazonia are closely related. However, the numerical relationship between satellite derived hot spot counts and the spatially coincident deforestation estimates at any spatial scale remains mostly unresolved. Among the major factors limiting our capacity to establish a more accurate relationship between hot spot counts and deforestation area are the following: 1. Vegetation fires have a highly dynamic nature. Constant changes in fire size and temperature limit our ability to derive a mean fire property. 2. The mode of image acquisition is noncontinuous. Satellite images are usually acquired at intervals ranging from 15 min to 12 h for geostationary and polar orbiting satellites, respectively. 3. The imaging process limitations involved. Optically thick clouds can obscure fires and prevent their detection [Schroeder et al., 2008a]. 4. The forms of burning and fire type variations. The wide range of vegetation structures and fuel loads which characterize Amazonia will influence the detection offires accordingly [Schroeder et al., 2005]. Due to the limiting factors above, it is very likely that a significant fraction of the actual fires will have only a few observations made by most remote sensing products during the entire life cycle of the burning event. Consequently, the relationship between the total deforested area and the number of hot spots detected for a particular location is usu-

" 1.:1.

r... !

.

1

ally difficult to derive. Nevertheless, the spatial distribution and concentration of hot spots tend to follow the trend described by the deforestation rates (Figure 2). Figure 3 shows the percentage distribution of hot spots detected alongside a 260-km segment of highway BR-163 near Novo Progresso in Para state. Fire detection statistics were derived for two distinct periods (1999-2000 and 200~ 2005) using seven lO-km buffers across the road's main axis. The 70-km-across subregion described by the buffers represents one ofthe most intact tracts of forests found in the immediate vicinity of highway BR-163 during this period. Factors such as reduced road trafficability, especially in the wet season, and the increased distance to ports and markets have limited the expansion of human activities in this region relative to other areas. For comparison purposes, the corresponding percentage dish'ibution of the annual deforestation increment derived from higher-resolution Landsat Enhanced Thematic Mapper Plus (ETM+) imagery is also plotted in Figure 3. Fire and deforestation show very similar patterns for the two periods analyzed with equivalent changes in the spatial distribution over time. Most important in Figure 3 is the progress in deforestation and fire use away from the highway and deeper into the forested areas, which suggests the intrusion of human activities in previously undisturbed areas. Absolute deforestation rates increased by a factor of three within the 5-year period analyzed, while hot spot counts went up by as much as five times. It is important to note that active fire detection products based on contextual methods, such as the one used with Figure 3, can be affected by commission errors which might reinforce the relationship with deforestation (see Giglio et al. [1999] for a discussion of different types of fire detection methods). These errors may be observed over deforested sites surrounded by relatively homogeneous forests as a result of the high thermal contrast between the two areas which cause a false detection to be produced [Schroeder et al., 2008b]. Despite the good overall agreement between the two different data sets in Figure 3, the measure of correlation describing individual episodes (i.e., the relationship between the number of hot spots detected and the area in hectares of the overlapping deforestation polygon) remains low (,2 = 0.54 using 2004 data). There has been a long debate over the relationship between roads, deforestation, and fire use in Amazonia [see, for example, Nepstad et al., 2001; Laurance et al., 2001; Silveira, 2001; Camara et al., 2005]. Roads facilitate access to otherwise remote areas and therefore serve to promote land use expansion where deforestation and fires play a significant role. However, their importance in relation to other forces such as regional and global economic markets is still subject to major controversy. Nevertheless, as shown in Figure 3,

~

SCHROEDER ET AL.

J .J: ~

~

lioooo

,

)1

14000

70000

"

12000

.."

, 60000

10000

50000

1

~

g-

8000

~

~ 5,000 head) in key Amazonian locations to traditional southern ranching districts. Their findings establish that Amazonian producers enjoy a higher rate of return on investments than their southern counterparts, due to low land prices and the resource base, with its abundant rainfall and solar energy. This advantage holds even though product prices in the south ate 10% to 20% higher given locations close to the prime markets. Although Amazonian soils are mainly Latosols with low fertility and high acidity, the fact remains that they are not much worse than those found in other cattle-producing parts of the country [Falesi, 1976; Fearnside, 1980; Adamoli et al., 1985]. Good moisture conditions, high insolation, and lack of frost compensate for soil limitations and allow for growth rates higher than elsewhere in the country [Margulis, 2004; Arima et al., 2005; Anualpec, 2003; Arima and Uhl, 1997]. Depending on type of operation (calving, fattening, etc.), large ranches in Amazonia enjoy 10-16% higher rates of animal growth, which translates into 5 to 10 additional kilograms of meat produced per hectare per year than on southern ranches (58-82 kg live growth ha- I versus 53-74 kg live growth per ha- I ). Such productivity advantages translate into higher profits, given low land prices [Arima et al., 2005; Anualpec, 2003; Barros, 2002]. The Arima et al. [2005] study documents land price differentials between north and south [see also Sawyer, 2008]. In Tupa, Sao Paulo, one of the traditional centers of southern production, they are nearly three times higher than in many important Amazonian production sites; here, a hectare of land goes for R$3300 compared to about $R1250 in Amazonia [Arima et al., 2005; Barros, 2002].Given ranches range into the tens of thousands of hectares, such a price differential represents a very large difference in capital costs reaching into the millions ofD.S. dollars for individual operations [A rima and Uhl, 1997].

With productivity advantages and lower costs for land, ranchers in Amazonia enjoy a higher profit rate than anywhere else in Brazil, despite the remoteness oftheir locations from major southern markets, and the impact of transportation costs on the prices they receive. The Tupa site yields a 4% internal rate of return (IRR); this does not compare well to Amazonian production, which yields an IRR nearly three times higher, at 12%. Evidently, Tupa is impacted by land costs that reflect the significantly higher rents obtainable from intensive agriculture [Arima et al., 2005], but is not an outlier among other southern locations in terms of low profit potential. Aggregate municipal calculations show an Amazonian return on investment at 5%, exceeding the average over all other states at 3.37% [Arima et al., 2005; Anualpec, 2003]. This reflects the fact that ranching is profitable across the board, for larg~ and small operations alike [Arima et al., 2005; Topall, 199~].

i 5.2. The Market Situation I

The profitability of Amazonian ranching is likely to translate into sectoral growth as markets continue expanding. Market expansion, in tum, will be driven by changes in supply and demand. As for the supply side, the dramatic insertion of Brazilian cattle 'products into the global market place is, in large part, a result of focused efforts to eradicate FMD. Recent outbreaks in Mato Grosso do SuI, Para, and Amazonas have dampened the initial upward spiral of Brazilian export, and 49 importing countries now impose certain restrictions on Brazilian products. Nevertheless, over the midrun exports are likely to surge again given that the infections appear to be the result of poor application of sanitary procedures rather than a new strain of the disease. Trade restrictions have generally not applied to all Brazilian states, and those unaffected by outbreak are continuing to export [Arima et al., 2005]. Since room for herd expansion is limited in cattle countries, such as Australia and Argentina, and since fear of bovine spongiform encephalopathy (BSE), or "I11ad cow" disease, restricts U.S. trade, Brazil, with its "undeveloped" lands, is the only country poised to supply any significant growth in world demand [Arima et al., 2005].

68

WALKER ET AL.

EXPANSION OF INTENSIVE AGRICULTURE AND RANCHING

As it turns out, these markets can be expected to expand such protection, so once the subsidies are removed, they will significantly over the next few decades. As is well-known, find their goods quite favorably priced [Arima et al., 2005]. meat is a "superior" good in economic terms, which means The fact that Brazilian cattle are mainly range-fed will make that as incljIDes rise, consumers tend to eat more of it. Thus, their meats even more desirable, since concerns continue to extensive scope exists for growth of demand given increas- linger about BSE. Given the role that Amazonian production ing incomes in China, and in Brazil itself, countries that will plays in both national and international markets, and given add to the population of consumers searching for meat in su- the profitability of Amazonian ranching, this will no doubt permarkets around the world. In China, yearly consumption translate into considerable expansion of the region's herd. of beef is 4 kg per person on average, which compares to 44 kg per person in the United States [U.s. Department of 5.3. The Dynamics ofthe Amazonian Herd Agriculture, 1997]. Clearly, the Chinese will ultimately exThe evolving economic environment faced by ranchers in perience some of their economic boom at the dinner table. the Amazon basin has enabled their steady march across the Further intensifYing the potential demand for Brazilian region's wild areas. Table 1 shows this by state, for the years beef is the weakening of political support for agricultural 1990 and 2005. In 1990, the Amazonian herd of 18 milsubsidies in developed countries, which keeps their prices arlion animals was widely dispersed, and four states already tificially low and competitive. Brazilian producers enjoy no

possessed counts exceeding a million head (Mato Grosso, Para, Rondonia, and Tocantins). By 2005, the pattern had dispersed further, a~ ranching occupied the far-flung corners of the basin. ,~oreover, the regional herd had reached 74,000,000, a sto~ larger than found in most cattle-producing countries. By M05, Acre joined the four other states supporting over a million animals, and Roraima, the only state without cattle in 1990, had a herd exceeding 500,000. Figures 1 and 2 graphically illustrate herd sizes throughout Amazonia Legal for the years in question (1990 and 2005). As can be seen, the early distribution suggests an advance starting in cerrado areas, with sizeable herds found in the southern and the eastern parts of the basin. A nearly continuous arc ofmunicipalities supported cattle herds, ranging from Paragominas in the northeast (para State), down through the south of Para, Tocantins and Mato Grosso, then west into Rondonia. The

Legend (In Heads)

Legend (In Heads) c:=J No data or No Production

D

No data or No Production

. . 1-100,000 _ 100,001 - 200,000

_

1 - 100,000 100,001 - 200,000

_

300,001 - 400,000

_

400,001 - 500,000 > 500,000

_

_

200,001 - 300,000

Source: IBGE, Pesquisa Pecuaria Municipal

_

300,001 - 400,000

o

_

400,001 - 500,000 > 500,000

200,001 - 300,000

_

legacy ranching on the natural grasslands of Maraj6 Island is observable, as are several anomalous locations beyond the initial cattle frontier, such as Itaituba in central Para, Jurua in northern Mato Grosso, and Rio Branco in Acre. The current distribution of the Amazonian herd shows a consolidation and advance of the frontier. As can be seen in Figure 2, the cattle arc is now completely continuous,joining up municipalities from Para all the way to Acre, thousands of kilometers to the southwest. In addition, the cattle-producing area now contains practically the entire state ofMato Grosso, as well as a sizeable portion of Para, such that nearly half the basin shows appreciable production. The only large remaining area with a few animals is in Amazonas, although Roraima and Amapa have yet to develop significant herds. Nevertheless, ranching has jumped the Amazon River course in western Para state, and cattle now forage in Monte Alegre,

Cattle by Municipality The Legal Amazon - 2005

Cattle by municipality The Legal Amazon -1990

250

I

Figure 1. Cattle herd by municipio, 1990.

500 I

1,000 Kilometers I

_

69

Source: IBGE, Pesquisa Pecuaria Municipal

o

250

I

Figure 2. Cattle herd by municipio, 2005.

500 I

1,000 Kilometers

I

70

EXPANSION OF INTENSIVE AGRICULTURE AND RANCHING

Alenquer, 6bidos, and Oriximina, to the north of Santarem. The dynamism of the Amazonian cattle economy is met by mechanized agriculture, our next topic of discussion. ~

6. THE EXPANSION OF SOY FARMING Like ranching, agriculture also has a long history in Amazonia: and its first riches were created, in part, by farming in the lower basin, where Portuguese colonists used slaves to cultivate or extract cocoa, coffee, cotton, and sugar cane [Santos, 1980]. That said, the current shape of farm production is a far cry from its early antecedents, and soy cultivation, in particular, represents a dramatic addition to Amazonian agriculture. Although the region produces many different crops, soy dominates by far in terms of revenues, yielding a 2005 crop valued at 7.8 billion reais, or about 61 % of the gross value of the harvest of annual crops in the region. Thus, it is a little wonder that soy, produced by highly capitalized production systems modernized far beyond what has traditionally passed as Amazonian farming, has come to symbolize the advancing edge of mechanized agriculture. The Amazonian soy boom of the past decade comprises an important part of national growth overall, which has expanded from an annual production of 20 million t in 1990, to 50 million t in 2004. This has led to an increase in soybean area planted from 115,847 to 215,972 km 2 in all Brazil. Such increases have made Brazil, the harvester of 28% of the global soybean crop, the world's second largest producer and exporter, supplying 27% of the world total. In comparison, the United States, the world's largest producer, supplies about 35% [USDA-FAS, 2004; IBGE, 2005]. Although traditional soy-producing states, such as Rio Grande do SuI, Mato Grosso do SuI, Sao Paulo, Minas Gerais, and Santa Catarina, supplied 54% of Brazil's total production in 1990, they had lowered their share to 37% by 1999. Currently, 33% of Brazilian soybeans are harvested in Amazonia Legal [CONAB, 2003; IBGE, 2005], a number that has grown dramatically from 24% in 1998. Evidently, this redistribution of production indicates a migration of soy production to the north, and the incorporation of the cheap, vast lands of the cerrado found there [Castro et al., 2001]. 6.1. The Supply Side: Infrastructure and New Cultivars

As with ranching, the boom of industrial agriculture in northern Brazil has been stimulated by auspicious changes in supply and demand, the latter ofwhich is linked to domestic and global markets. On the supply side, infrastructure investments have promoted soy production by bringing about a precipitous decline in transportation costs. As has already been discussed, the development of road networks has cut

days off travel times between northern and southern destinations in Brazil. The ports, waterways, and rails developed under various federal initiatives have also been especially important for Amazonian soy production. The establishment of major transshipment points on the Rio Madeira in Porto Velho, and a deep water port on the Amazon River itself, in Santarem, has decreased transportation costs for soybean produced in the upper basin. In eastern Amazonia, the Ferrovia Norte-SuI (north-south railway) now provides a rail link to connect soy producers in Maranhao and Tocantins to the Port of Itaqui on the Atlantic coast. Besides infrastructure expansion, the recent development of soybean cultivars suitable for hot humid conditions has proven decisive. Soybeans are naturally short-day plants adapted for growth in subtropical and temperate areas. Thus, as with ranching, early thinking about soy farming in Amazonia emphasized environmental limitations, and a popular and scientific consensus emerged that climatic conditions would ultimately inhibit the development of a robust soy economy above latitude 25° [McGrath and Vera-Diaz, 2006; Jordan, 1982; Sioli, 1973]. Adding further support to this general view was the fact that Brazilian soybeans were originally cultivated with great success between 20 0 S and 300 S, where U.S. cultivars were well adapted to the local climate and soil [EMBRAPA-SOJA, 2002]. Nevertheless, genetic modifications have opened the door to Amazonian production, and the current expansion into low latitudes is possible with new cultivars possessing long-juvenile genes, which delay flowering and maturity. Without long-juvenile genes, soybean plants grown at low-latitudes flower too soon; this makes them short and difficult to harvest mechanically [Hartwig and Kiihl, 1979; Sinclair et al., 2005; VeraDiaz et al., 2008]. 6.2. Growing Demand

As for growth in world demand, this has, in large part, been sparked by robust economic expansion in China, and the ri.sing global consumption of vegetable oils and soyconsuming poultry, swine, and livestock. Today an estimated 30% of the world's vegetable oil consumption and 70% of protein meal consumption are derived from soy (Soystats at www.soystats.com). which provides an effective and sanitary substitute for animal parts, the primary vector of disease transmission in animal-fattening operations [Rohter, 2003; Vera-Diaz et al., 2008]. Between 1990 and 2000, soybean global demand grew 68% (from 104.2 to 175.2 million t), and global consumption of soybean grain increased 41 % (from 104 to 146.7 million t) [AGRIANUAL, 2000; RCW, 2004]. During this same period, Brazil expanded its share of the global market from 15% to 22% (15.4 to 38.4 million t)

WALKER ET AL.

-

i

and exported 64%. of its 'production. In addition, domesti c soybean consumptiOn doubled, from 6.6 to 13.6 million 1. .A num.ber of stud~p6 suggest that global demand for soy Will contmue to]rw o.ver the next few decades. As with beef, many soy p ducts are considered "superior" goods and eve~ conse - ative predictions from the United Nation; PopulatIOn Division [2004] and RCW [2004] suggest that the . demand for soybeans will increase from 225 .6 ml'11'Ion t m 2001 to 385 million tin 2020. Future Brazilian dynamics have also been a~dressed, considering production and export figures ~bserved m th~ 1990s. In particular, Brazil's soybean productIon could easily - grow to 73 million t by 2020 ,M'th ~ore than 55 million t exported to global markets [Rodrigues, 2004].

71

ticularly Roraima, where production, nonexistent in 1998, reached 36,000 thy 2005. Only Acre and Amapa appear to ?e unaffected by the soy boom, although output remains low m Amazonas (5136 t). . T~ese spatial dynamics are shown for a longer time period m Figures 3 and 4 (1990 to 2005). Here can be observed the advance of the soy frontier, first in the central and southern cerrados of Mato Grosso, where it is highly concentrated in 1990: Although the some soy was produced in Tocantins and m Maranhao, as well as in the border areas between Mato Grosso and Rondonia, Amazonian soy fanning in the early 1990s. was largely a single-state phenomenon. This changes rad~cal~y b~ 2005 (Figure 4). Clearly, the majority ~f soy fa~mg IS stIll found in Mato Grosso, with productIon practIcally everywhere in the state excepting areas in 6.3. The Dynamics o/Soy Production the bordering Rondonia"Para and az oAm nas. D far. northwest . esplte ItS growth and concentration in Mato Grosso soy . The explosive growth of soybean production in Amazo' ma from 3 to 2.0 million t a-I between 1990 and 2005 has has taken root throughout the Amazon basin. As for Mato Grosso, the figure reveals a substantial westbeen accomp~med by increases in area planted from 16,000 ward ;n?vemen~of farming into the southeastern parts of to 70:000 ~ [IBGE, 2005]. Disaggregated state dynamics Ro.ndoma, c~eatIng, with Mato Grosso, a nearly continuous ~re given m Table 3 for the period from 1998 to 2005 durStriP of soy m southern Amazonia Legal, buffered only b mg which Pro~uctio~ became a region-wide pheJ1om~non. we~lands to the south. In addition, several widely disperse~ Soy had estabhshed Itself by 1998 in both Mato Grosso and fOCI hav_e emerged. ~rom the east in the states of Bahia and Rondonia, although Mato Grosso clearly dominated with an Maranhao, soy fa~mg now merges into the significantly output of about 7 million t. By 2005, the regional pattern had expan~ed pro~uctlOn zone of Tocantins. These areas, in ch~nged. Soy farmers in Mato Grosso significantly increased With th:!;: Mato Grosso croplands. The graphical tum, Imk up their output to nearly 18 million t, but Rondonia and Para t~e soy frontier could overrun the native data suggest that both ex.ceeded 200,000 t of production, with output in Para cerrados of the basip., which occupy its southern and eastern expand.mg al~ost ten times over the 7-year period. Further, flanks [see Mueller, 2003]. Tocantms, With a reasonable production in 1998 (123,085 t), Besid~s .Mato Grosso, Rondonia, and Tocantins, Para bec~~e the second largest Amazonian producer, with nearly sho~s slgmficant emergent production, with rather high out1 milhon.tons (905,328). States producing less than 100,000 t puts m Paragominas and Santarem. Incipient soy farming is by 2005 mclude Acre, Amapa, Amazonas and Roraima Of observable, along the Transamazon Highway (e.g., Altamira these, Roraima and Amazonas showed s;ong growfu, ~ar~nd Uruara) an~ now forms a corridor along BR-163 linkmg the ~rod~ctlOn areas in Mato Grosso with Santarem. Soy, farmmg IS also found in five municipios in the south of Table 3. Growth in Soy Production Amazonian States' Pa.ra_(Santana do ~raguaia, Santa Maria das Barreiras, ConAve Annual ce19ao do AragUa/a, Redenyao, and Floresta do Araguaia) 1998 States Growth (% a-I) a~d across the Amazon River from Santarem, in Alenquer 2005 300 Acre Fmally, the cerrados to the far north in Roraima now sup~ 114 -0.0 Amapa 0 port a. crop, as do the Amazonas municipios across the Rio Amazonas 0 0.0 68 796 5136 Madeira from Porto Velho (Humaita) and in close proximity Mato Grosso 7228052 17 ,761,444 " 18 to Manaus.

Para Rondonia Roraima Tocantins

2438 , 15 ,970 0 123,085

204,302 233,281 36,400 905,328

1034 172 NAb 79

'V~I~es given in tons. Data source: IBGE, ProdUl,;ao Agricola MUlliclpal (PAM). ~A, not available.

7. IMPLICATIONS FOR THE AMAZONIAN LANDSCAPE .Agri.cultural development is of great importance to Brazil gIVen ItS comparative natural advantages. A long growing season and cheap land have conspired to make it a world

72

EXPANSION OF INTENSIVE AGRICULTURE AND RANCHING

WALKER ET AL.

73

Soybean in the Legal Amazon Planted Area - 2005 "

Legend (In Hectares)

Legend (In Hectares)

CJ No data or No Production

CJ No data or No Production

_

1 - 250 251-500

_

1 - 250 251 - 500

_

501 -1,000

_

501 -1,000

_

1,001 - 5,000

_

5,001 - 10,000

_

10,001 - 25,000

_ _

25,001 - 50,000 >50,000

source: IBGE, Producao Agricola Municipal (PAM)

o I

255

510 I

1.020 Kilometers

I

Figure 3. Planted soy area by municipio, 1990.

powerhouse in the production of agricultural commodities. But agriculture possesses a land-demanding production function, which means that when it expands into forested areas, trees give way to crops and pasture grasses over large areas. The expansion of pasture into Amazonia has long generated controversy in this regard, and it is a fact that pasture constitutes the lion's share of cleared lands in the north region [Walker et al., 2008]. The question is: Can we expect such agriculturally driven clearance to continue, and if so, by how much? Earlier sections of this chapter addressed the demand picture for both cattle products and soy. Projections of impacts on the Amazonian landscape require that this picture be distilled into an estimate of the demand for land.

7.1. Agricultural Expansion and Von Thiinen

In tackling such a challenge, it is useful to situate Amazonian agriculture in a conceptual framework that links commodity demands to the input of land for production. To this end , we consider the model of von Thiinen, who pointed out that (l) agricultural activity occurs so long as rents are positive, and that (2) rents are functions of prices for the products of land and inputs to their production. Von Thiinen also noted that landscapes reveal spatial patterns of crop locations with intensive forms of land use found in nearer areas of ~opulation concentration, and extensive ones, farther away.

_

1,001 - 5,000

_

5,001 - 10,000

-

10,001 - 25,000

_

25,001 - 50,000 > 50,000

source: IBGE, Producao Agricola Municipal (PAM)

o I

255

510

I

1,020 Kilometers

I

Figure 4. Planted soy area by municipio, 2005.

With the Thunian framework, we conceptualize the expansion of the agricultural frontier into Amazonia as being driven by increasing rents [Walker, 2004; Walker et al., 2008]. These rents are bolstered by prices for farm and ranch products stemming from globalizing demand, and by production cost reductions due to improvements in the transportation system [Mueller, 2003]. At the lead edge is ranching, found far from population centers because it generates rents far from market centers. Behind ranching comes the complementary advance of soy and mechanized agriculture. more generally. Under a Thunian formulation, deforestation is the manifestation of an advancing agricultural frontier, occurring when potential rents, previously nonexistent by virtue of market or infrastructure conditions, become positive [Walker and Solecki, 2004].

Two issues must be addressed before we consider Amazonian land cover dynamics within this conceptual framework. The first is that of agricultural intensification. Simply put, intensification is the adoption of new farming practices or technologies that raise output per unit land. Consequently, intensification leads to reduced demands for land, ceteris paribus, and for this reason, many have appealed to it as a solution to the problem of deforestation. The second issue involves the mechanisms of forest loss under a multicrop Thunian system and specifically the role of soy expansion in driving Amazonian deforestation, given that cattle ranching is an active partner. As for intensification, it is often imposed on a farmer or farming group by virtue of land scarcity, as has been extensively observed in the historic record [Boserup, 1969]. For

74

EXPANSION OF INTENSIVE AGRICULTURE AND RANCHING

the Amazonian case, fanners and ranchers are unlikely to intensify production by much, given the abundance of cheap land, even with new technologies. Moreover, with abundant land, inten~ve systems can promote deforestation if they generate more rents than the nonintensive system [White et al., 2001; Arima et al., 2005]. Thus, in the following discussion, we do not consider intensification per se and address the case of technologically static systems for both ranching and soy production. In addition, we do not regard the replacement of pasture by say as a fonn of technological intensification, strictly speaking, and reserve the tenn from this point on to describe increasing efficiency ofland use for the production of specific crops, such as soy itself. Thus, soy fanning intensifies if newly implemented technologies yield higher, unit-area soy production. The second issue involves the identification of the underlying forces driving an advanced edge of agriculture or ranching into so-called "uncultivated wilderness," to use the original tenninology of von Thiinen [Walker, 1999]. In a single commodity world, as for example with simply ranching, such an exercise devolves to describing the circumstances that would increase rents for ranching outputs such as beef. If prices for meat rose, for example, areas would be brought into production that previously did not generate positive rents. A similar result is obtained with a reduction in transportation costs. In a two-commodity world, the picture grows more complicated. Assuming that, to begin with, soy is found "behind" the cattle frontier; several possibilities arise. The first is that market conditions change for only one ofthe products, soy or beef. If the price for beef rises but not for soy, then we are in the situation first described, with an advancing cattle frontier as has been historically observed over the past several decades. On the other hand, if the price for soy rises but not for beef, soy advances into areas that were previously pastured, until such time as available pastures are exhausted, and only forest land remains available. At that point, soy directly replaces forest. . The circumstances as described are unrealistic for the Amazonian case, as it appears that the market situation of the recent past has generally favored both soy and ranching [Brandao et al., 2005]; in addition, the benefits of infrastructure do not discriminate by crop, and all agricultural activities on a frontier receive a rent windfall with new investments. Thus, soy potentially affects Amazonian forest cover by two mechanisms. In the first instance, soy may be "pushing" the cattle frontier deeper into forest [Sawyer, 2008; Vera-Cruz et al., 2008]. This occurs if soy occupies productive pasture land, due to rising rents, and if rents continue to rise for cattle products. In the second instance, market conditions may be stronger for soy, in which case soy production "leapfrogs" into areas of primary forest, in advance of ranching.

Land use leapfrogging, driven by the same mechanism, is observed when sprawl swallows up agriculture on the urban periphery and converts natural areas into residential land use [Walker and Solecki, 2004].

WALKERET AL.

!ab!e 4. Deforested Area Converted to Mechanized Agriculture and Area Converted to Mechanized Agriculture From Forest Reported III Literature for Three Study Areas

,

~)1

"

Study Location

Time Period Analyzed

Deforested Area" (k:m2) (a)

Mato Grosso state

2001-2004

38,097

Santarem and Be1terra, Para municipios

1975-1986 1986-1997 1997-1999 1999-2004 2004-2005 1996-2001

7.2. The Greater Impact: Pasture or Soy?

As has been discussed, the expansion ofpasture has been a prime driver of deforestation in the Amazon basin. The question that now arises is, "What impact will the Amazon's new big crop, soy, have on the forest?" We speculate that soy has exerted both "pushing" and "leapfrogging" effects, alluded to in the conceptual discussion above, but that market conditions favoring soy are currently emerging. We base this on remote-sensing analysis of forest conversion, resolved into type of land use, and size of clearing. Data on the fate of deforested lands are not available for the entire Amazon basin, but studies have used remote sensing to distinguish among pasture, mechanized cropland, mainly soy, and other land uses in three areas, the state of Mato Grosso [Morton et ai., 2006, 2009a], Santarem and Belterra municipalities in the state of Para [Venturieri et ai., 2007], and Vilhena municipality, in the southeastern part of Rondonia [Brown et ai., 2005]. In this regard, Morton et al. [2006] report that 12% to 14% of deforested area in Mato Grosso was converted directly to cropland between 2001 and 2004 (Table 4). The percentage peaks at 23% in 2003 when both deforestation rates and the price of soybeans were relatively high, which is consistent with a Thunian "leapfrog" of soy into the primary forests of Mato Grosso. In addition, land deforested for cropland accounted for 28% to 33% of all lands converted to mechanized agriculture. This counters the claim that all cropland expansion occurs only on previously cleared lands, mainly pasture [Morton et al., 2009a, 2009b]. A similar pattern is observed in Vilhena in eastern Rondonia. Although the majority of cropland expansion takes place on previously cleared land, a substantial portion (22% from dense forest and 200/0 from less dense forest) directly consumed forested land between 1996 and 2001 [Brown et ai., 2005]. Taking a longer-tenn perspective, Venturieri et al. [2007] document the emergence of mechanized agriculture as a driver of deforestation since 1975 in the municipalities of Santarem and Belterra, Para. No mechanized soy production was present in the study area prior to 1999, but in the periods 1999-2004 and 2004-2005,8% and 2.7% of new croplands were created from forests, respectively. The study area also exhibits a smaller proportion of direct conversion to mechanized cropland than in Mato Grosso or Rondonia (8.2% and 10.7% for the two periods). In Para, denser forest may favor the use of already-cleared pasture by expanding croplands.

75

Vilhena, Rondonia municipio

. Deforested Area Conve11ed to Area Converted Mechanized to Mechanized Agriculture Agriculture b Area (krn2) (b) (c) 16,370

4670-5463

Deforested Area Converted to Mechanized Agriculture (% of total) (c)/(a)

Area Converted to Mechanized Agriculture From Forest (% of total) (c)/(b)

12.25-14.34

28.53-33.37

Source

Morton

et at. 821 739 419 527 140 not reported

0.0 0.0 0.0 544 560 70.36

0.00 0.00 0.00 44 15 15.71 (dense) 14.21 (less dense)

0.00 0.00 0.00 8.35 10.71 not reported

0.00 0.00 0.00 8.09 2.68 22 (dense) 20 (less dense)

[2006] Venturieri et ai. [2007]

Brown et ai. [2005]

:Includes parcels deforested for all use~ includ~ng mechanized agriculture, pasture, and not yet in production. Includes parcels converted to mechanIzed agnculture from all land covers including forests, pastures and successional vegetation.

Data on size of clearings associated with individual deforestation events also points to the growing importance of soy and mechanized agriculture, more generally, as a direct cause of forest loss in the Amazon basin. In particular, mechanized operations involving soy typically make large clearings quickly in order to hasten the start-up of production, given the presumed degree of capitalization and risks associated with soy fanning. By way of contrast, a relatively larger component of pasture creation is linked to small and medium producers, with low levels of technology, who proceed in a piecemeal fashion in fonning their pastures over longer periods of time [Walker, 2003].

Table 5 indicates that the vast majority of annual clearings were less than 100 ha in 2001-2005, which reflects the large populations of smallholders in the basin who mainly clear the forest to make Way for pastures; in which case, they represent the advanceq edge of the cattle frontier [see Walker, 2003; Walker et al.,:2008]. Medium (100-1000 ha) and large (> 1000 ha) clearings, dedicated to both pasture and soy, are relatively few in number, but they contribute disproportionately to area cleared. Thus, medium-sized clearings account for only 3% of all clearings, but 37% of the region's deforestation; large clearings, at less than 1%, account for 13% of forest cleared. These proportions vary widely by state, with

TableS.. Proportion of Number of Deforested Polygons and Deforested Area in Small «100 ha), Medium (100-1 OOOha) and Lar e g Categones (> 1000 ha) by State for 2001-2005 Number of Deforested Polygons in Size Category (%)

Deforested Area in Size Category (%)

State

Deforested Area (2001-2005)" (krn 2)

Small

Medium

Large

Small

Medium

Large

Acre Amapa Amazonas Maranh1io b Mato Grosso Para Rondonia Roraima Tocantins Total

3,344 111 5,120 4,642 44,959 33,840 16,427 1,312 985 110,743

0.99 0.99 0.99 0.97 0.92 0.96 0.97 0.99 0.98 0.94

0.01 0.01 0.01 0.03 0.08 0.03 0.03 0.01 0.02 0.03

rule on private holdIngs _ __=50:%_ __..----5J5~O~~~~~~~~~~~~K~1I0~m~e~teJrs 70'fJO'W

89

f;lYffO"",,,

Figure 2. Deforestation by 2020, scenario 2: Expected population growth, Avanya Brasil investments, partial governance.

:100,0 -Rivers -Highways # Cities State Capitals ! A Country Capitals ~ Study Areas E'Za Rural Settlements I::::: :1 Indigenous, Conservation, and State and Federal Protected Areas

V)

p

V)

p

0

0

S

S

V) V)

P0 p,

P 0 P,

V) V)

P

p

Gross Income (R$ (% From Farming))

lEGE Size Classes

FNO

Gross Income (R$)

Total Area (ha) 90

2,126.6

500-1,000

0

'"

27,500

5-10 10-20 20-50 50-100

::

28

-9

£

u

'E :::J

.c u

t;:::

'0 Q)

Cl.

2 91"

1+

~

00

22 20

o

o

. '0 . ~ .0

700

'00'00'0,' 0 "

600

0

\;

500

(J)

u c ro

18

0

16

'0 ~

-

14

~

12

'l"V.~ "V'v. \ "\. I

Q)

NT

..v;~

l':'y.

....

\~ r:rf.

v

J

.rv~\VT..

\}? " V '. / ~

Cl.

E

2

~

E

g

fi

'. Vl

."\: ¥?.

'\

::J

167

400

"

.,

T

v\rr"" v \ 'trv.vv

E 5 c 0

~

300

:t=

200

0..

Cl.

'0 Q)

L-

100

10 0 Jan/99

Jul/99

Jan/OO

Jul/OO

Jan/01

Jul/01

Jan/02

Jul/02

Jan/03

800 600

'0 (f)

.S:

400

'--

())

co 200

+-'

:s:

'+--

0

())

800

(c)

2.0 - 3.4 m

-e..- Pasture

0>

~

0

-.0.-

600

Forest

C /)

400 200 Jan/99

Jul/99

Jan/OO

Jul/OO

Jan/01

Jul/01

Jan/02

Jul/02

Jan/03

Figure 4. (a) Monthly averages of air temperature (circles), specific humidity (inverted triangies), and monthly totals of precipitation (columns) measured over the forest (represented by solid symbols) and pasture (represented by open symbols) from FeblUaty 1999 until September 2002. Storage of water in the soil at forest and pasture sites for the layers fi'om (b) 0-2 m and (c) 2-3.4 m deep. Adapted from von Randow et at. [2004], reprinted with permission of Springer-Verlag.

ence dominates the difference in Sn between forest and pasture. Outgoing L n has a large seasonal variation, peaking in August when the atmosphere is d1y (Figure 2b) and cloud cover is a minimum. Outgoing L n is slightly larger over the pasture primarily because of warmer mean daytime temperatures. Throughout the year, Rn is greater over forest because of its lower albedo and slightly smaller outgoing long-wave flux. The mean al1l1ual reduction in Rn over the pasture is 13.3% [von Randow et al., 2004], smaller than the 20.4% given for the boreal forest by Betts et al. [2007], where snow cover also reduces R n over grass in late winter.

2.3. Diurnal Cycle ofthe Amazonian BL

The daytime convective BL over Amazonia is rarely cloud-free, so the depth of the mixed layer below cloud base, the lifting condensation level and the near-surface RH are all tightly coupled [see Betts et al., 2006]. Near-surface RH is strongly influenced by the availability of water for evaporation, so forest sites where rooting is deep (Figure 4) and pasture'sites show larger differences in the d1y season than in the wet season. Afternoon mixed layer heights range from 700 to 1100 m in the rainy season over forest and pasture, when

168

AMAZONIAN BOUNDARY LAYER AND MESOSCALE CIRCULATIONS

Figure 5. Monthly averages of (top) surface albedo and (bottom) net short-wave radiation (Sn, circles), net long-wave radiation (L n, squares) and net all-wave radiation (R n, inverted triangles) over forest (solid symbols) and pasture (open symbols) during 1999-2002. Adapted from von Randow et al. [2004], reprinted with permission of Springer-Verlag,

Bowen ratios are low and RH is high [Fisch et al., 2004; von Randow et al., 2004]. In the dry season, strong subsidence (Figure 2a) brings dty air into the BL, and evaporation is reduced over the pasture in Rondonia, so mixed layer depths are much larger and can reach 2000 m over the pasture. Betts et al. [2002a] and Strong et al. [2005] discuss the surface diurnal cycles of temperature, humidity, lifting condensation level, equivalent potential temperature, surface fluxes, and BL cloud for easterly and westerly regimes (see section 3.3) at the Rondonia pasture site in the 1999 rainy season. They show that the downward solar radiation and the fluxes of sensible and latent heat are lower for the westerly wind regime, which has more stratiform cloud but has a higher water vapor mixing ratio with a weaker diurnal cycle. The easterly wind regime shows an early morning maximum of mixing ratio, followed by a fall to a minimum in the afternoon, as the cumulus clouds mix water vapor up and out

of the subcloud layer more rapidly than is provided by surface evaporation. As the rainy season progresses throughout January and February 1999, there is a steady transition toward cloudier conditions and lower surface fluxes. Daytime surface Bowen ratio for this pasture site is about 0.4 and falls slightly as the rainy season progresses. Typically, in the afternoon, evaporatively driven downdrafts from convective rainbands transfOlm the boundaty layer. The fall of equivalent potential temperature in the boundary layer is about 10K and is similar for both regimes, but the boundaty layer cooling by individual convective events during the westerly regimes is reduced because the subcloud layer is shallower on average, as rain events are weaker but more frequent. This boundary layer modification by rainbands is rather similar to that seen in other moist convection regimes in the tropics (e.g., in Venezuela [Betts, 1976]). Figure 6 compares the seasonal cycle of the diurnal cycle for the near-surface variables for the Rondonia pasture and forest sites during 2001, using data from von Randow et aI, [2004]. Note that the temperature, T, and relative humidity data RH are fi'om Vaisala HMP35A instruments, mounted at v~ry different heights: 8.3 m over the pasture and 60 m over the forest floor, well above the mean canopy height of 35 m [von Randow et al., 2004]. The "wet" season mean is January, February, and March; the "dry" season here is simply August, when mean subsidence is strongest over Rondonia; and the "dry-to-wet" transition is an average of September and October. Although the mean temperature in Figure 6a varies little over the year, the diurnal amplitude of temperature doubles between the wet and dty seasons, as the atmosphere gets drier and less cloudy, and the outgoing net long-wave radiation, which is a primaty driver of the diurnal temperature range [Betts, 2006], doubles (see Figure 5). The forest data (at 60 m) are warmer than the pasture (at 8.3 m) in the rainy season, but the pasture becomes watmer in the daytime in the dty season with a greater increase in diurnal range. The diurnal moisture structure in Figure 6b shows the seasonal fall of water vapor mixing ratio, Q, between wet and dty seasons and the rapid recovety by the transition season. The forest always has a greater mixing ratio than the pasture [von Randow et al., 2004]. For Figure 6, a low bias of 4.3% near saturation at the pasture site was corrected, but it is possible that part of the humidity difference between sites is still instrumental, as these instruments are only calibrated to a few percent in RH. Indeed, absolute calibration of humidity instruments is difficult in these very moist tropical environments [Betts et aI" 2002c]. Away from the rainy season, mean mixing ratio rises steeply from a morning minimum at sunrise, when the atmosphere is saturated at the surface (Figure 6d), as evaporation is trapped in the stable nocturnal BL.

BETTS ET AL. 34 - - r - - - - - - - - - - - - - - , Pasture: Season

19

-;f-Ory:8

32

~j>-

-transition: 9,10

,"'---Wet: 1,2,3

18

Foresl: Season

,

~

Ol

.>