Identification of Eucalyptus Areas Based on Spatial and Temporal Segmentation of The Landsat Time Series

Abstract

The conversion of natural systems into anthropized systems has been causing an overload on ecosystems and changes in the landscape. Brazil is one of the biggest practitioners of forestry, but information about the sector is quite conflicting. It is necessary to understand how these changes occur and multitemporal Remote Sensing emerges as an analysis tool. The objective of the present work is to test a methodology for detecting eucalyptus areas using the LandTrendr algorithm on the Google Earth Engine platform combined with Object Oriented Analysis from the Landsat satellite time series, between 1985 and 2020, and to identify the age of the segments. found using the same. The con fusion matrix showed an overall accuracy of 0.990 with the Area² algorithm and a Kappa of 0.959, presenting a very satisfactory result. The identification of the mode of the first year of gain of each segment obtained a Kappa of 0.643.