Fitomonitoração e modelagem de fotossíntese em jatobá (Hymenaea courbaril L.) com redes neurais artificiais. / Phytomonitoring and modelling of photosynthesis in jatobá (Hymenaea courbaril L.) with artificial neural.

AUTOR(ES)
DATA DE PUBLICAÇÃO

2003

RESUMO

The increases in greenhouse gas concentrations, mainly carbon dioxide, and the climatic changes have become important scientific, economic, and political subjects in the past years. The Kyoto Protocol establishes the Clean Development Mechanism, which grants carbon credits for projects that promote the sequestration of carbon in developing countries. Therefore, it is important to evaluate the CO2 absorption capacity by terrestrial plants, and this requires the development of gas flow and gas exchange models in different scales. That development is usually complicated, because the ecophysiological processes are non-linear. This work presents a method to model photosynthesis at the leaf level, as a first step toward quantifying the potential of carbon sequestration. The technique used was artificial neural networks (ANNs), as it allows the adjustment of non-linear relationships between input and output variables. The work was divided in two parts: phytomonitoring and modeling. The phytomonitoring was accomplished in jatoba (Hymenaea courbaril) during one year. The following physiologic variables were measured: photosynthesis rate, transpiration rate, stomatal conductance, leaf temperature, and fluorescence; and environmental variables: CO2 concentration, photosynthetic activity radiation, relative humidity, and air temperature. An unprecedented amount of data for that type of experiment and for that plant species was obtained. The analysis of these data showed important characteristics about the relationship of the physiologic variables in Hymenaea courbaril and the environmental variables, in the four seasons. The data collected were used for the modeling and fine-tuning of the neural network. The network was trained with different combinations of input variables to observe to which group of variables the neural network responded better. The analysis of the training results showed that with the ANN technique it is possible to achieve a very good approximation of the photosynthesis function, with 92% success rate for entries consisting of filtered data.

ASSUNTO(S)

phytomonitoring efeito estufa modelagem photosyntesis fitomonitoração neural networks redes neurais greenhouse effect modelling jatobá carbon sequestration fotossíntese jatobá seqüestro de carbono

ACESSO AO ARTIGO

Documentos Relacionados