天美传媒

Abstract

The relationship between biodiversity loss and the impacts on ecosystem services of tropical forests, in face of the ongoing global climate change needs to be better quantified. In this work, we considered the concept of Ecosystem Integrity (IE), which represents the connection of biodiversity with the ability of ecosystems to sustain the processes of self organization. Bayesian Networks (BBN-Bayesian Belief Network) can provide metrics for the generation of ecosystem integrity index, from the training of probabilistic relationships of evidence obtained through field data, Remote sensing data and GIS. The objective of this work is to present the methodological approach and the results of IE mapping, elaborated at the regional scale for different patterns of phyto-ecologic landscape of the Brazilian Amazon. The modeling was based on learning from the parameters (datadriven model) through the use of the Expectation Maximization algorithm. For the validation of this probabilistic model, an evaluation was carried out in controlled areas, with field observation and comparison with the IE model based on knowledge (knowledge driven), prepared by experts.

Biography

Margareth Simoes is a senior researcher at Embrapa where is the coordinator of several Projects on Geomatics for Environmental and Agriculture Planning. From 2010-2012 she was a researcher at Embrapa LabEx Europe Program, where was responsible for the research area of Agriculture Sustainability and Natural Resources, coordinating projects and international articulations in Europe. By that time, she was a research fellow at la Maison de la T茅l茅d茅tection (UMR TETIS Territoires, Environnement, T茅l茅d茅tection et Information Spatiale). She is also professor at the Department of Computer Engineering and thesis supervisor at the Post graduate (Ph.D.) Course on Environment at Rio de Janeiro State University, Brazil.