Multi-temporal analysis of radiometric changes in satellite images of forest edges to infer selective-logging areas in the Amazon forest

• Autor(a) principal: GRAÇA,Paulo Maurício Lima de Alencastro
• Data de Publicação: 2015
• Outros Autores: MALDONADO,Francisco Dario, SANTOS,João Roberto dos, KEIZER,Edwin Willem Hermanus
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Acta Amazonica
• Texto Completo: http://old.scielo.br/scielo.php?script=sci_arttext&pid=S0044-59672015000100035
• topic: timber exploitation; deforestation; remote sensing; forest monitoring

Radiometric changes observed in multi-temporal optical satellite images have an important role in efforts to characterize selective-logging areas. The aim of this study was to analyze the multi-temporal behavior of spectral-mixture responses in satellite images in simulated selective-logging areas in the Amazon forest, considering red/near-infrared spectral relationships. Forest edges were used to infer the selective-logging infrastructure using differently oriented edges in the transition between forest and deforested areas in satellite images. TM/Landsat-5 images acquired at three dates with different solar-illumination geometries were used in this analysis. The method assumed that the radiometric responses between forest with selective-logging effects and forest edges in contact with recent clear-cuts are related. The spatial frequency attributes of red/near infrared bands for edge areas were analyzed. Analysis of dispersion diagrams showed two groups of pixels that represent selective-logging areas. The attributes for size and radiometric distance representing these two groups were related to solar-elevation angle. The results suggest that detection of timber exploitation areas is limited because of the complexity of the selective-logging radiometric response. Thus, the accuracy of detecting selective logging can be influenced by the solar-elevation angle at the time of image acquisition. We conclude that images with lower solar-elevation angles are less reliable for delineation of selecting logging.