Sensoriamento remoto para a modelagem da biomassa e biodiversidade arbórea em Minas Gerais: contexto temporal e espacial

• Autor(a) principal: Pereira, Jhuly Ely Santos
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/46317
• topic: Sensoriamento remoto; Biodiversidade florestal; Manejo florestal; Biomassa vegetal; Remote sensing; Forest biodiversity; Forest management; Vegetable biomass; Recursos Florestais e Engenharia Florestal

Monitoring vegetation over large territorial extensions is essential to define management strategies for ecosystem services conservation, and remote sensing by satellite is a valuable tool for this task. In this sense, it is essential to develop modeling methods based on remote sensing variables that provide reliable estimates of vegetation parameters, such as aboveground biomass and tree biodiversity, quickly and at low cost. In this work, the predictive performance of random forest models, based on spatial and temporal variables derived from the enhanced vegetation index (EVI) and the land surface temperature (LST), was evaluated to estimate the aboveground biomass (AGB) and the tree species diversity (TSD) in the tropical forests of Minas Gerais, Brazil. This dissertation is divided into two parts. In the first part (General Introduction), we tried to situate the reader in front of the research objectives, making a theoretical approach on the themes worked. In the second part, two articles were presented. In article 1 (Annual Indices of Remote Sensing for Modeling Aboveground Biomass and Biodiversity in Minas Gerais), images from the MODIS sensor were used to model AGB and TSD based on the time variation over the year in the values of EVI and LST. In Article 2 (Textural Metrics for Modeling Aboveground Biomass and Biodiversity in Minas Gerais), TM sensor images were used on board Landsat 5 to model AGB and TSD from the spatial variation of EVI and LST in the studied areas. In general, in both articles, the results indicate that the performance of the prediction models is mainly affected by the degree of complexity of the vegetation structure.