Predição do sequestro do carbono em área reflorestada utilizando redes neurais artificiais

Carregando...
Imagem de Miniatura

Data

2014-02-07

Autores

Nunes, Clodoaldo [UNESP]

Título da Revista

ISSN da Revista

Título de Volume

Editor

Universidade Estadual Paulista (Unesp)

Resumo

O aumento das emissões e concentrações de gás carbônico na atmosfera tem ocasionado a intensificação do efeito estufa e, consequentemente, das alterações climáticas e desequilíbrio sobre o planeta. Deste modo, torna-se imprescindível a adoção de medidas e o desenvolvimento de métodos eficientes para quantificar e avaliar a emissão do fluxo de gás carbono na atmosfera pela vegetação terrestre. Assim sendo, esta pesquisa tem por objetivo propor o desenvolvimento de um sistema inteligente, baseado no emprego de redes neurais, para predizer , seja a curto, médio ou de longo prazo, o sequestro (absorção) de carbono em áreas de reflorestamento, em uma fazenda situada na Amazônia Mato-Grossense com 10 mil hectares. O sistema proposto foi desenvolvido utilizando uma composição formada por uma rede neural da família ART (Adaptive Resonance Theory), a ART-Fuzzy e uma rede neural feedforward multicamadas com treinamento realizado via uso do algoritmo retropropagação, quantificando e realizando a previsão com maior acurácia e precisão da absorção de carbono, colaborando, desta forma, para que se adotem medidas para minimizar os impactos futuros do aumento da temperatura global e garantir a sustentabilidade do planeta. Ressalta-se que esta proposta é inovadora, pois pode-se abordar de áreas com diversidade de árvores, de grande extensões plantadas, entre outros fatores importantes neste contexto. Ou seja, trata-se de um sistema inteligente robusto e de grande flexibilidade de aplicação.
The increase in emissions and concentrations of carbon dioxides in the atmosphere has brought the intensification of greenhouse effect and, consequently, climatic changes and imbalance on the planet. Therefore, it becomes essential to adopt measures and develop efficient methods for quantifying and evaluating the emission of flow of carbon in the atmosphere by land vegetation. Thus, this research aims to propose the development of an intelligent system, based on the use of neural networks to predict, whether in the short, medium or long term, the carbon sequestration in reforestation areas, on a 10000 hectare farm located in the Amazon region of Mato Grosso-Brazil. The proposed system was developed using a composition formed by a neural network of the ART family (Adaptive Resonance Theory), the Fuzzy-ART and a multilayer feedforward neural network with training performed through the use of the retropropagation algorithm, making the quantifying and predicting of the carbon absorption, thus contributing to the adoption of measures to minimize the impacts of future global temperature increase and ensure the sustainability of the planet, more accurate and precise than the original methods. It is highlighted that this proposal is innovative, because it can approach large areas of plantations, with diversity of trees, among other important factors in this context. That is, it is an intelligent system robust and of great flexibility of application.

Descrição

Palavras-chave

Carbono, Reflorestamento, Redes neurais (Computação), Sustentabilidade, Carbon

Como citar

NUNES, Clodoaldo. Predição do sequestro do carbono em área reflorestada utilizando redes neurais artificiais. 2014. 94 f. Tese (doutorado) - Universidade Estadual Paulista Júlio de Mesquita Filho, Faculdade de Engenharia de Ilha Solteira, 2014.