TIME SERIES FOREST CHANGE DETECTION USING LANDTRENDR: STUDY OF CASE IN RIO DE JANEIRO STATE

Abstract

Understanding how terrestrial systems evolve is important in pursuing strategies that optimize the use of natural resources and minimize environmental impacts. Monitoring vegetation cover and land use changes through remote sensing techniques has been crucial in this regard. The main objective of this paper was to contribute methodologically to the monitoring of terrestrial systems through remote sensing techniques, using all the available collection of Landsat images, in an approach based on time series. For this trajectory-based change detection was used through the LandTrendr algorithm to detect changes in forest cover in Rio de Janeiro's state between 1984 and 2016, identifying different types of disturbances (deforestation and recovery) and classifying age of secondary forests. In a general way observed of 58969 hectares of forest changes in the state of Rio de Janeiro, these are 64% of deforestation and 36% of regenerations. This map achieved 70% global accuracy. On the coast, deforestation is old and more abrupt, while inland forest losses are also old but gradual, creating a constant degradation of the landscape. Forest recoveries in the Rio de Janeiro's state are more recurrent in Lagos Tourist region. A synthesis map was developed with trajectory characteristics. Only the Lagos region has forest recovery tendencies, whereas the regions in the interior of the state have tendencies of degradation. The regions near the coastal have tendencies of stability which rates of gain and loss very similar.