Adriana Aparecida Moreira

EVALUATION OF REMOTELY SENSED EVAPOTRANSPIRATION PRODUCTS MOD16 AND GLEAM IN EDDY COVARIANCE FLUX SITES FROM LBA PROJECT

This study aims to evaluate the performance of MOD16 and GLEAM evapotranspiration (ET) datasets in nine eddy covariance monitoring sites. Data from both ET products were downloaded and its daily means calculated. Evapotranspiration estimations were then compared to the observed ET in the eddy covariance monitoring sites from the Large-Scale Biosphere-Atmosphere Experiment in the Amazon (LBA). We performed a statistical analysis using the correlation coefficient (R), the root mean square error (RMSE) and BIAS. Results indicate that, in general, both products can represent the observed ET in the eddy covariance flux towers. MOD16 and GLEAM showed similar values to the calculated statistics when ET estimates were compared to observed ET. Model estimates and eddy covariance flux towers are subject to uncertainties that influence the analysis of remotely-sensed ET products.

O Algoritmo Support Vector Machine Aplicado ao Mapeamento do Uso e Ocupação do Solo (The Support Vector Machine Algorithm Applied to Mapping and Land Use)

As rapidas transformacoes do meio ambiente decorrentes das atividades humanas de ocupacao dos espacos tem se tornado, uma das grandes preocupacoes atuais. Nesse contexto, torna-se fator essencial a analise do uso e cobertura terrestre e seu monitoramento a fim subsidiar informacoes para gestao ambiental. Atraves de mapeamentos realizados com a utilizacao de ferramentas de Sensoriamento Remoto e possivel a obtencao de informacoes acerca das mudancas de uso e cobertura da Terra, sendo este realizado por meio de tecnicas de classificacao de imagens, utilizando algoritmos classificadores. Dentre os algoritmos desenvolvidos para este fim, existem aqueles, baseados na area do conhecimento da Inteligencia Artificial como o Support Vector Machine (SVM), que vem sendo empregado com sucesso na separacao maxima das classes. O objetivo deste trabalho e a aplicacao do SVM no mapeamento do uso e cobertura do solo na Bacia do Rio Vieira e comparar os resultados com aqueles obtidos pelo algoritmo, Maxima Verossimilhanca. Nos resultados encontrados, observou-se que na analise do indice Kappa ambos os algoritmos apresentaram uma forte concordância e as classes analisadas se mostraram coerentes com as caracteristicas da regiao de estudo, contudo o algoritmo SVM apresentou menor confusao espectral entre as classes e melhor desempenho operacional. A B S T R A C T The rapid changes in the environment resulting from human activities of occupation of spaces have become one of the major current concerns. In this context, it becomes essential analysis of use and land cover and its monitoring in order subsidize information for environmental management. Through mappings performed with the use of tools for Remote Sensing is possible to obtain information about changes in Land use and cover, this being accomplished by means of techniques of classification of images, using algorithms classifiers. Among the algorithms developed for this purpose, there are those, based in the area of knowledge of Artificial Intelligence such as the Support Vector Machine (SVM), which has been employed with success in maximum separation of categories. The objective of this work is the application of SVM in the use and coverage of the soil in The Basin of the River Vieira and compares the results with those obtained by the algorithm, Maximum Likelihood. In the results found, it was observed that in the analysis of the Kappa index both algorithms presented a strong concordance and the classes analyzed if showed consistent with the characteristics of the region of study, however, the SVM algorithm showed lower spectral confusion between categories and better operational performance. Keywords : Artificial Intelligence, River Basin Vieira and Classification