Detecção da mortalidade de bambus em florestas do sudoeste amazônico usando dados MODIS (MAIAC)

The southwest Amazon forests are dominated of bamboo species of the genus Guadua and cover about 161,500 km2. These individuals have a life cycle of about 28 years and present only one reproductive event followed by their death. This process accumulates dry biomass in the forest and increases the probability of fire occurrence. Since extreme drought events increase fire vulnerability, and this region has been affected by some of these events in the last decades, it is possible that bamboos are related to a higher occurrence of forest fires. In this sense, the objective of the present study was to evaluate the detection of bamboo mortality in southwest Amazon forests using MODIS (MAIAC) data, spectral mixture modeling and CART decision tree. The combined use of band 5 (medium infrared) and GV fraction allowed the detection of bamboo mortality with 56% accuracy. The bamboo life stages can be differentiated because of different spectral patterns, especially on the near and medium infrared wavelengths, and distinct intra-annual seasonality patterns from non-bamboo forest. It was possible to map the bamboo-dominated forests, including bamboo life stages, and non-bamboo-dominated forests with 76% accuracy. The mortality detected in this study may have been influenced by the occurrence of fire.