The forests in the Brazilian Amazon have undergone heavy pressures in recent decades not only due to deforestation but also from degradation processes (e.g. unsustainable logging, forest fires, etc.), which are still poorly monitored. The Brazilian National Institute of Spatial Research (INPE) has mapped forest degradation in the Brazilian Amazon (DEGRAD project, INPE 2008) since 2007 but a preliminary assessment showed that many selectively logged forest areas are not being included.
Mapping selective logging is still challenging because (1) logging evidence (e.g. roads, logging decks, skids) can be detected by remote sensing imagery only for a limited amount of time because the forest canopy closes rapidly after the logging event (Asner et al., 2009); (2) logging result in a complex environmental landscape including vegetation, dead trees, bare soil, etc. (Souza, 2009) and (3) selective logging and forest fires happen in a synergistic and recurrent way, showing highly dynamic and complex temporal and spatial patterns (Souza, 2013). Landsat imagery has been extensively used for mapping selective logging in the Brazilian Amazon (Asner et al., 2005; Souza et al. al, 2005, Matricardi et al. 2007, 2012). However, Landsat datasets have limitations especially due to the spatial resolution.
The objective of our study is to compare Landsat 8 and Sentinel-2 imagery for mapping selective logging in the Brazilian Amazon focusing on the assessment of the benefits of combining the two types of imagery for such purpose. Moreover depending upon Sentinel-2 image availability, we expect to perform an assessment of intra-annual changes in order to better understand selective logging spatial-temporal patterns.