Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Brésil → Uso de dados do sensor MSI/Sentinel-2 e de LiDAR aerotransportado para mapeamento de fitofisionomias de caatinga e estudo das relações com atributos físico-químicos dos solos

Instituto Nacional de Pesquisas Espaciais – INPE (2018)

Uso de dados do sensor MSI/Sentinel-2 e de LiDAR aerotransportado para mapeamento de fitofisionomias de caatinga e estudo das relações com atributos físico-químicos dos solos

Silveira Hilton Luís Ferraz da

Titre : Uso de dados do sensor MSI/Sentinel-2 e de LiDAR aerotransportado para mapeamento de fitofisionomias de caatinga e estudo das relações com atributos físico-químicos dos solos

Auteur : Silveira Hilton Luís Ferraz da

Université de soutenance : Instituto Nacional de Pesquisas Espaciais – INPE

Grade : Mestrado do Curso de Pós-Graduação em Sensoriamento Remoto 2018

Résumé
Caatinga is a natural semi-arid vegetation type, which occupies great part of northeastern region of Brazil. This ecosystem contains a variety of phytophysiognomies of difficult mapping with their occurrence influenced by local rainfall and soil attributes. This work verified the potential use of multi-temporal data from the MSI/Sentinel-2, obtained in four dates between 2015 and 2016 (rainy to dry seasons), along with LiDAR observations, for mapping seven Caatinga´s phytophysiognomies in a study area located in the state of Pernambuco. Using a vegetation reference map and Random Forest (RF) classification, eventual gains in classification accuracy have been evaluated from multi-temporal over mono-temporal MSI data (rainy and dry seasons) ; from adding vegetation indices into the analyses ; and from inserting LiDAR metrics into the classification. The relationships between the mapped vegetation by RF and 20 physico-chemical attributes of 75 soil profiles were studied by using principal component analysis (PCA) and ordinary kriging. The results showed that : (a) there were no differences in classification accuracy between the dry and rainy seasons ; (b) multi-temporal data improved classification accuracy compared to mono-temporal observations ; (c) a smaller number of vegetation indices had similar classification performance than a greater number of reflectance of bands ; and (d) LiDAR metrics improved classification accuracy of arboreous and sub-shrub classes (11,1% and 10,7% respectively). Statistically significant differences were observed in organic carbon content, cation exchange capacity, water retention at field capacity, horizon thickness, soil porosity and rock fragments (% pebble, gravel, fine-earth fraction) between these two types of phytophysiognomies (arboreous and sub-shrub classes)

Version intégrale

Page publiée le 3 avril 2019