Selective Logging in the Brazilian Amazon

See allHide authors and affiliations

Science  21 Oct 2005:
Vol. 310, Issue 5747, pp. 480-482
DOI: 10.1126/science.1118051


Amazon deforestation has been measured by remote sensing for three decades. In comparison, selective logging has been mostly invisible to satellites. We developed a large-scale, high-resolution, automated remote-sensing analysis of selective logging in the top five timber-producing states of the Brazilian Amazon. Logged areas ranged from 12,075 to 19,823 square kilometers per year (±14%) between 1999 and 2002, equivalent to 60 to 123% of previously reported deforestation area. Up to 1200 square kilometers per year of logging were observed on conservation lands. Each year, 27 million to 50 million cubic meters of wood were extracted, and a gross flux of ∼0.1 billion metric tons of carbon was destined for release to the atmosphere by logging.

Tropical forests have been threatened by increasing rates of deforestation or clear cutting during the past three or more decades (1). Although deforestation, largely for the conversion of land to food crops or pastures, is the major destructive force in tropical forests worldwide (2), other forest disturbances such as the selective harvest of timber have also increased in frequency and extent (3, 4). In selective logging, a limited number of marketable tree species are cut, and logs are transported off site to sawmills. Unlike deforestation, which is readily observed from satellites, selective logging in the Brazilian Amazon causes a spatially diffuse thinning of large trees, which is hard to monitor by using satellite observations. Selective logging causes widespread collateral damage to remaining trees, subcanopy vegetation, and soils; with impacts on hydrological processes, erosion, fire, carbon storage, and plant and animal species (310).

There is little known about the extent or impacts of selective logging throughout the tropical forests of the world, including the Amazon Basin. A survey of sawmills in the Brazilian Amazon suggested that 9,000 to 15,000 km2 of forest had been logged in 1996–1997 (3). The large uncertainty in this reported area resulted from necessary assumptions of the wood volume harvested per area of forest. Sawmill surveys can, at best, provide only a general idea of where and how much logging occurs, because most operators buy timber at the mill gate rather than harvesting the wood themselves.

Objective spatially explicit reporting on selective logging requires either labor-intensive field surveys in frontier and often violently contested areas or remote detection and monitoring approaches. Previous studies of small areas show the need for high-resolution observations via satellite (1113). Moreover, most of the traditional analysis techniques used for localized selective-logging studies have been insufficient for large-scale selective-logging assessments (11, 14, 15). A detailed comparison of Landsat satellite observations against field measurements of canopy damage after selective logging proved that traditional analytical methods missed about 50% of the canopy damage caused by timber harvest operations (16).

We advanced the computational analysis of Landsat Enhanced Thematic Mapper Plus (ETM+) satellite data using the new Carnegie Landsat Analysis System (CLAS) to detect and quantify the amount of selective logging in the major timber-production states of the Brazilian Amazon. This approach provides automated image analysis using atmospheric modeling; detection of forest canopy openings, surface debris, and bare soil exposed by forest disturbances; and pattern-recognition techniques. CLAS provides detailed measurements of forest-canopy damage at a spatial resolution of 30 m × 30 m, and it does so over millions of square kilometers of forest (17).

We applied CLAS to five states—Pará, Mato Grosso, Rondônia, Roraima, and Acre (fig. S6)—which account for ∼90% of all deforestation in the Brazilian Amazon (18). The analysis was conducted on a time series of Landsat ETM+ imagery from 1999 to 2002. Across the five timber-producing Brazilian states, the annual extent of selective logging ranged from 12,075 to 19,823 km2 (Fig. 1). These logging results represent new forest damage not accounted for in deforestation studies. Each year, the overlap between our results and the Brazilian National Institute for Space Research (INPE) annual deforestation maps was only 6% (±5%) (17) (Fig. 2). Moreover, only 19% (±11%) of the total area logged in any given year was subsequently deforested 3 years later. Selective logging thus adds 60 to 123% more forest-area damage than has been reported for deforestation alone in the same study period (Table 1). Selective logging was concentrated in the states of Mato Grosso and Pará, where logging areas exceeded or nearly matched deforestation areas. In other smaller states, selective logging increased forest damage area by 10 to 35% over reported deforestation rates (Table 1).

Fig. 1.

Spatial distribution of selective logging in five timber-production states of the Brazilian Amazon for the year intervals 1999–2000 (red), 2000–2001 (blue), and 2001–2002 (green). The states of Amazonas (AM), Amapa (AP), Tocantins (TO), Maranhao (MA), and the southern non-forested part of Mato Grosso were not included in the analysis. Light gray areas show the extent of indigenous reserves; dark gray areas delineate federal conservation lands as of 1999 (29). RR, Roraima; PA, Pará; MT, Mato Grosso; RO, Rondônia; AC, Acre.

Fig. 2.

High-resolution example of selective-logging results in 2001–2002 from the CLAS processing (right), compared with deforestation mapping provided by the INPE (left) (18).

Table 1.

Selective-logging rates from 1999–2002 in five major timber-producing states of the Brazilian Amazon, with comparison to the deforestation rates reported by the INPE (18).

State 1999-2000 rates (km2 year-1) 2000-2001 rates (km2 year-1) 2001-2002 rates (km2 year-1)
Logged Deforested Logged Deforested Logged Deforested
Acre 64 547 53 419 111 727
Mato GrossoView inline 13,015 6,176 7,878 7,504 7,207 6,880
Pará 5,939 6,671 5,343 5,237 3,791 8,697
Rondônia 773 2,465 923 2,673 946 3,605
Roraima 32 253 55 345 20 54
Total 19,823 16,112 14,252 16,178 12,075 19,963
  • View inline* Only the northern 58% of Mato Grosso containing forested lands was included in the analysis.

  • Conservation units such as indigenous reserves, parks, and national forests generally afforded protection against logging. However, exceptions included areas in northern Mato Grosso, where up to 53, 291, and 50 km2 of logging were measured each year in the Xingu, Aripuanã, and Serra Morena indigenous reserves, respectively (Fig. 1). In the southern portion of Pará state, major logging disturbances were observed in the Menkragnoti and Kayapó indigenous reserves, with up to 261 and 198 km2, respectively, detected each year between 1999 and 2002. Federal forest reserves of Acre, Gorotire (Pará), and Juruena (Mato Grosso) were harvested for timber at rates of up to 23, 90, and 380 km2 each year, respectively.

    Extensive field validation studies showed that the detection of canopy damage within CLAS is precise and accurate (17). Field validation studies showed false-positive and false-negative detection rates of only 5%. Uncertainty caused by errors in atmospheric correction of satellite data, cloud cover, annualization, automated logging-area delineation, and manual auditing were 0.7 to 12.8%, individually. After combining all known sources of error, our analysis suggests an overall absolute uncertainty of up to 14% in total logging area.

    Selective logging contributes substantially to gross carbon fluxes from the Brazilian Amazon. We combined forest-damage results from CLAS with field-based forest-canopy gap-fraction (19, 20) and roundwood-extraction data (21) to calculate the total wood-extraction rates (17). In 2000, 2001, and 2002, roundwood production averaged 49.8, 29.8, and 26.6 million m3, respectively. The mean annual harvest intensities were 26.6, 21.7, and 21.4 m3 ha–1, which were generally lower than those reported by sawmill owners in 1996 (3, 22). The total volume harvested equates to 10 million to 15 million metric tons of carbon (MtC) removed (23). In addition to roundwood, residual stumps, branches, foliage, and roots are left to decompose in the forest, subsequently returning to the atmosphere as carbon dioxide over about a decade. Our calculated average harvest intensity of 23.2 m3 ha–1 is equivalent to ∼8 MgC ha–1 contained in roundwood, with an associated 34 to 50 MgC ha–1 of fine and coarse debris (2325). Integrated to the regional scale, the processing of roundwood and the decomposition of residues lead ultimately to a gross flux of carbon from the forest of up to 0.08 GtC for each year of logging (26). This value increases the estimated gross annual anthropogenic flux of carbon from Amazon forests by up to 25% over carbon losses from deforestation alone (27). Post-harvest forest regeneration reduces the net flux of carbon to the atmosphere below these values, but the pace of regeneration after logging varies considerably (9, 28).

    Selective logging doubles previous estimates of the total amount of forest degraded by human activities (Table 1), a result with potentially far-reaching implications for the ecology of the Amazon forest and the sustainability of the human enterprise in the region. In the future, improved monitoring of tropical forests will require high-performance satellite observations and new computational techniques. Our results, presented with explicit uncertainty analysis and transparency of method (17), have located and quantified ubiquitous but previously cryptic disturbances caused by selective logging.

    Supporting Online Material

    Materials and Methods

    Figs. S1 to S7

    Tables S1 to S5

    References and Notes

    References and Notes

    Stay Connected to Science

    Navigate This Article