Detalhes bibliográficos
| Autor(a) principal: |
Ferraz, Antonio Santana |
| Data de Publicação: |
2012 |
| Tipo de documento: |
Tese
|
| Idioma: |
por |
| Título da fonte: |
LOCUS Repositório Institucional da UFV |
| Texto Completo: |
http://locus.ufv.br/handle/123456789/575
|
Resumo: |
Estimates of carbon and biomass in forested areas can be obtained using allometric equations. High resolution remote sensing imagery has also been widely used to estimate tree carbon biomass, based on digital variables (radiance/ reflectance values and vegetation indices) extracted from images. In natural forests, that are typically heterogeneous and present high floristic, physiognomical and phenological diversity, the use of this technique is more complex since there is little quantitative biomass data collected in the field and a lack of research that integrates data from different sources such as forest inventories and satellite images to obtain estimates. In this study a set of IKONOS II satellite images was used to estimate aerial biomass and carbon stock in a semideciduous seasonal forest fragment located in Viçosa, MG, in an area known as Mata da Silvicultura , belonging to the Universidade Federal de Viçosa. Estimates of above ground biomass and carbon stocks were obtained with allometric equations based on forest inventory data conducted in fifteen 1,000 m2 (20 m × 50 m) parcels. These estimates were related to digital variables (reflectance of four spectral bands and 12 vegetation indices) extracted from IKONOS II satellite images using regression analysis and an artificial neural network. In regression analysis the highest (significant) correlations to carbon stock and biomass were found for the spectral bands 2 and 4 and the GEMI, SAVI, TCap1, TCap2 and TCap3 vegetation indices. However, only variables Band4 and TCap1 were necessary for estimating both total biomass and carbon stock. For biomass estimates, correlation coefficients of r2 = 0.394 for Band4, and r2 = 0.496 for TCap1 were observed. For carbon stock estimates, correlations of r2 = 0.400 with Band4 and r2 = 0.504 for TCap1 were found. Through neural network training it was found that use of the four IKONOS II satellite spectral bands as input variables were sufficient to estimate biomass and carbon stocks for the area studied. Residuals obtained using regression analysis exceeded 60% while residuals in the range of ±1.5% were obtained using neural networks. |