Uso da reflectância de imagens Landsat 5 TM na identificação de plantios de Eucalyptus dunnii e Eucalyptus urograndis e sua correlação com o volume de madeira

• Ano de defesa: 2014
• Autor(a) principal: Goergen, Laura Camila de Godoy
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; BR; Recursos Florestais e Engenharia Florestal; UFSM; Programa de Pós-Graduação em Engenharia Florestal
• Palavras-chave em Português: Sensoriamento remoto; Índice de vegetação; Inventário florestal; Remote sensing; Vegetation index; Forest inventory; CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL
• Link de acesso: http://repositorio.ufsm.br/handle/1/8730

The objective of this study was to test the potential of satellite imagery, TM/Landsat 5, for discrimination of plantations of different ages of Eucalyptus dunnii and Eucalyptus urograndis and correlate the volume of these plantations, obtained from forest inventory, with the spectral responses. The values of spectral reflectance of the surface of the original images were recovered and after image geocoding the values of reflectance were extracted in six spectral bands TM sensor (B1, B2, B3, B4, B5 and B7) stand for the four groups studied: E. dunnii age 3 and age 5 and E. urograndis to 2.2 years and 4.2 years of age. In addition to the spectral bands vegetation indices SR, NDVI, SAVI_0.5, SAVI_0.25, MVI and GNDVI were used. To evaluate the behavior of the spectral variables for each stand, it was performed an analysis of principal components which, for the year 2009 , the variables B2 , B3 , GNDVI , B4 , B5 and B1 , were, in descending order , the most significaqnt. And for the year 2011, the most significant values were the SAVI_0.25, SAVI_0.5, B4, SR, MVI, NDVI and B2 variables, in descending order. From the discriminant analysis data of three discriminant functions (λ) to separate the four groups were generated. The structural attributes with better discriminatory power (in order of importance) were: SAVI_0.25, SAVI_0.5, B5, MVI, B7, B1 and B3. The discriminant model generated showed that functions correctly classified 100% of the cases in their predicted groups, revealing that the spectral variables were good predictors for distinguishing plantations. Correlation analysis between the biophysical variable (timber volume) was not significant for the planting of E. dunnii at 3 years old. For the planting of E. dunnii at 5 years was the most correlated variable B2 (r= -0.55). The B4 was the variable most strongly correlated with the volume in plantations of E. urograndis at 2.2 years old (r= 0.75) followed by the index Ln (SAVI_0.5) with r= 0.72. For E. urograndis at 4.2 years of age, the variables with the highest correlation were B2 (r= 0.67), followed by Ln (SAVI_0.5) with r= 0.63. From the correlation coefficients obtained, equations to estimate the volume were modeled. For the settlement of E. dunnii at 5 years, the best fitted equation explained 48% of the variability in the volume. The population of E. urograndis at 2.2 years obtained the best results, in which 57% of the variability was explained by the volume of spectral variables. The population of E. urograndis at 4.2 years obtained the lowest results, where only 45% of the variability was explained by the volume spectral variables. It was concluded that the methodology can be used to aid in species identification from satellite images and further studies should be conducted to estimate volume from spectral variables.