Assessment of tropical forest degradation by selective logging and fire using Landsat imagery
Introduction
In recent decades, the logging of high-value tree species has became an important land use of tropical evergreen forests in Central Africa (Laporte et al. 2007) and in the Brazilian Amazon (Stone and Lefebvre, 1998, Nepstad et al., 1999, Alvarado and Sandberg, 2001). However, until the 1960s most of this logging in Brazil was restricted to flood plains (várzeas) in areas flooded annually. This situation changed drastically after the widespread construction of roads during the 1960s and 1970s which allowed the expansion of selective logging into the inter-fluvial (terra-firme) forests in the Brazilian Amazon (Uhl and Vieira, 1989, Uhl et al., 1997).
As a result of the increased access provided by roads, selective logging became a major concern in the Brazilian Amazon due to its potential negative effect on natural forests. Selective logging is a form of timber extraction of a group of trees from selected species where only the most valuable tree species are removed from the forest. Selective logging operations usually leave behind a complex landscape comprised of intact forest, treefall gaps, roads, log-loading patios, and damaged forest (Verissimo et al., 1995, Uhl et al., 1997, Laporte et al., 2007). In addition, logging activities increase the amount of dead slash or dried biomass (fuel) and, consequently, forest fire susceptibility is substantially increased (Uhl and Buschbacher, 1985, Stone and Lefebvre, 1998, Nepstad et al., 1999, Souza and Barreto, 2000). Therefore, the impacts caused by logging in tropical forests are significant both in terms of forest degradation and fire susceptibility (Uhl and Buschbacher, 1985, Stone and Lefebvre, 1998, Nepstad et al., 1999, Souza and Barreto, 2000). Selectively logged forests often become highly degraded and usually have 40–50% of their canopy cover removed during logging operations (Uhl and Vieira, 1989, Verissimo et al., 1992). Accordingly, forest fires have increased in the Brazilian Amazon, and became very common in fragmented forests located adjacent to deforested areas, which serve as ignition sources for forest fires (Uhl and Kauffman, 1990, Cochrane and Schulze, 1998, Cochrane, 2001, Cochrane et al., 2004).
Meanwhile, most studies based on remotely sensed data to estimate forest impacts of selective logging in the Brazilian Amazon have relied on study sites mostly located in Eastern Pará, Brazil. For example, Souza and Barreto (2000) used a linear mixture model and Landsat imagery to detect logging patios within selectively logged forests in a study site in the state of Pará, Brazil. Additionally, Cochrane and Souza Jr. (1998) developed a remote sensing technique to detect and classify burned forests using non-photosynthetic vegetation derived from Linear Mixture Analysis for a study site in Tailândia, Pará. Furthermore, Cochrane and Sousa, 1998, Cochrane et al., 1999 used field studies and a multi-temporal analysis of remotely sensed imagery to understand forest fire dynamics in another study site in Pará. Finally, Souza et al. (2003) developed a methodology to map classes of degraded forest for a case study in Pará using fraction images (vegetation, non-photosynthetic vegetation, soil, and shade) derived from spectral mixture models.
Asner et al. (2002) applied Landsat textural analysis to assess forest canopy damage from selective logging in yet another study site in the state of Para, Brazil. Based on those study results, the authors concluded that although this technique was useful for broad delineation of forests impacted by selective logging, it could not appropriately estimate the extent of canopy damage.
More recently, Souza et al. (2005a) conducted an evaluation of different vegetation and infrared indices and fraction images derived from spectral mixture analysis to assess multi-temporal forest degradation within nineteen transects in the eastern Brazilian Amazon, whilst Matricardi et al. (2005) estimated areas of selectively logged forests using Landsat imagery and fieldwork measurements for a study site in the Brazilian state of Mato Grosso. Souza et al. (2005b) developed a remote sensing approach based on Spectral Mixture Analysis (SMA) to map selectively logged and burned forests for a case study in the state of Mato Grosso.
Asner et al. (2005) also developed an automated remote sensing technique to map degraded forests by selective logging in the five major timber center states of the Brazilian Amazon. The authors used Landsat ETM + imagery from 1999 to 2002 and the Camegie Landsat Analysis (CLAS) based on atmospheric modeling and spatial pattern analysis to detect forests impacted by selective logging in their study area. By using the CLAS, they were able to classify forests degraded by selective logging on the Landsat imagery with 86% overall accuracy. That study did not include, however, any assessment of forest degradation by forest fires.
These previous studies have contributed to substantial improvements in remote sensing based techniques to assess both extent and impacts caused by selective logging and forest fires. However, a more comprehensive assessment of forest disturbances by forest fire and selective logging and their interactions with other land use and land cover processes is still lacking. Moreover, atmospheric changes caused by smoke from deforested and forested areas have limited multi-temporal analysis using remotely sensed data (Karnieli et al., 2001, De Moura and Galvão, 2003). This research intended to measure extent and to assess impacts of selective logging and forest fire on tropical rain forests using Landsat imagery. Therefore, we conducted a detailed multi-temporal analysis of these impacts on natural forests in the Southern Amazon State of Mato Grosso, one of the major logging centers in Brazil. It included field observations, statistical tests on performances of different vegetation indices and green vegetation fraction (GV) derived from SMA to assess forest canopy degradation in the presence of smoke, and development of a modified vegetation index more resistant to atmospheric conditions in the study site.
Section snippets
Study site
This research was conducted using one Landsat scene (path 226 and row 068) that encompassed approximately 30,000 km2 in the state of Mato Grosso, located in the southern Brazilian Amazon (Fig. 1). The study area included the municípios of Santa Carmem and União do Sul and parts of the municípios of Colider, Feliz Natal, Itaúba, Marcelândia, Nova Ubiratã, Paranatinga, Sinop, Sorriso, and Vera, which combined form a territory known as the Sinop region.
The climate in the study area is humid
Relationships among various vegetation indices
In this analysis, the relationship among NDVI, MSAVI, GEMI, AFRI, MSAVIaf, GEMI2.1, and GV retrieved from a Landsat image acquired on August 6, 2003 were tested under smoky and smoke-free atmospheric conditions. Table 5 illustrates the performance of various vegetation indices under smoke-free atmospheric condition in the study site.
This correlation matrix (Table 5) demonstrates high correlation (r ≥ 0.93) among all vegetation indices under clear sky atmospheric conditions. The highest
Conclusions
Based on these research results, fractional coverage derived from vegetation index domain and retrieved from Landsat TM and ETM + imagery, showed to be an accurate and efficient approach in measuring forest canopy degradation and regeneration in tropical evergreen forests. Both modified vegetation indices (AFRI and MSAVIaf) showed better performance than SMA, GEMI, and NDVI in predicting forest canopy cover area under smoky conditions. However, MSAVIaf proved to be more accurate than AFRI to
References (61)
- et al.
Classification of multispectral images based on fractions of endmembers: application to land-cover change in the Brazilian Amazon
Remote Sensing of Environment
(1995) Under the hood: issues in the specification and interpretation of spatial regression models.
Agricultural Economics
(2002)- et al.
Remote sensing of selective logging in Amazonia — assessing limitations based on detailed field observations, Landsat ETM + and textural analysis
Remote Sensing of Environment
(2002) - et al.
A method for manual endmember selection and spectral unmixing.
Remote Sensing of Environment
(1996) - et al.
Radiometric normalization of multitemporal high-resolution images with quality control for land cover change detection
Remote Sensing of Environment
(2002) - et al.
Radiometric rectification: toward a common radiometric response among multidate, multisensor images
Remote Sensing of Environment
(1991) A soil-adjust vegetation index (SAVI)
Remote Sensing of Environment
(1988)Sensitivity of the normalized difference vegetation index to subpixel canopy cover, soil albedo, and pixel scale
Remote Sensing of Environment
(1990)- et al.
AFRI — aerosol free vegetation index
Remote Sensing of Environment
(2001) Linear mixture modeling approach for estimating cotton canopy ground cover using satellite multispectral imagery
Remote Sensing of Environment
(2000)
Biophysical parameter estimations using multidirectional spectral measurements.
Remote Sensing of Environment
A modified soil adjusted vegetation index.
Remote Sensing of Environment
Spatial and temporal dynamics of vegetation in the San Pedro River basin area.
Agricultural and Forest Meteorology
Combining spectral and spatial information to map canopy damages from selective logging and forest fires
Remote Sensing of Environment.
Logging impacts and prospects for sustainable forest management in an old Amazonian frontier: the case of Paragominas.
Forest Ecology and Management
Extraction of a high-value natural resource in Amazonia: the case of mahogany.
Forest Ecology and Management
Soil moisture estimation in a semiarid rangeland using ERS-2 and TM imagery
Remote Sensing of Environment
Fire in the Brazilian Amazon: a spatially explicit model for policy impact analysis
Journal of Regional Science
Selective logging in the Brazilian Amazon
Science
A vegetation index which minimizes soil brightness effects on LAI and APAR estimation
Mapping Target Signatures via Partial Unmixing of AVIRIS Data
Revised Landsat-5 TM radiometric calibration procedures and postcalibration dynamic ranges.
IEEE Transactions on Geoscience and Remote Sensing
Roads, land use, and deforestation: a spatial model applied to Belize
The World Bank Economic Review
Synergistic interactions between habitat fragmentation and fire in evergreen tropical forests.
Conservation Biology
Positive feedbacks in the fire dynamic of closed canopy tropical forests.
Science
Forest fires in the Brazilian Amazon.
Conservation Biology
Selective logging, forest fragmentation, and fire disturbance
Linear mixture model classification of burned forests in the Eastern Amazon.
International Journal of Remote Sensing
Smoke effects on NDVI determination of savannah vegetation types.
International Journal of Remote Sensing
Cited by (129)
Mapping tropical forest degradation with deep learning and Planet NICFI data
2023, Remote Sensing of EnvironmentPost-logging changes in a neotropical dry forest composition and structure modify the ecosystem functioning
2023, Forest Ecology and ManagementForest structure and solar-induced fluorescence across intact and degraded forests in the Amazon
2022, Remote Sensing of EnvironmentMapping and recycling proposal for the construction and demolition waste generated in the Brazilian Amazon
2022, Resources, Conservation and RecyclingTimber production and carbon emission reductions through improved forest management and substitution of fossil fuels with wood biomass
2021, Resources, Conservation and RecyclingThe smokescreen of Russian protected areas
2021, Science of the Total EnvironmentCitation Excerpt :Forest degradation, logging and fire are well-known major threats for global forests (Matricardi et al., 2010; Burivalova et al., 2014; Cazzolla Gatti et al., 2015, 2019a; Cazzolla Gatti and Velichevskaya, 2020) and their impacts increase every year (Laurance et al., 2012; Vaglio Laurin et al., 2016; de Oliveira Roque et al., 2018).