Espectroscopia do Infravermelho Próximo (FT-NIR) como um atalho para estudos funcionais e taxonômicos em larga escala na Amazôni

Abstract

The aim of this research was to evaluate the Near Infrared Spectroscopy (FT-NIR) as a shortcut to large-scale ecological and botanical studies in the Amazon. In the first chapter, I asked if branch samples could be used to discriminate species better than leaves, as well as higher taxonomic levels, since leaves are generally more variable and plastic than wood. I used branch and leaf samples from Reserva Ducke, Manaus, and found results greater than 90% accuracy in the discrimination of species, genera and family. The results achieved with the use of branch branches were equated to leaf values. I conclude that Near Infrared Spectroscopy (FT-NIR) has great potential for discriminating species based on branch samples and is suitable for discriminating a diverse range of Amazonian tree genera and families. This opens possibilities for a variety of botanical and ecological studies in areas of high plant diversity. In the second chapter, I evaluated whether a regional model based on FT-NIR spectroscopy data would be efficient for estimating 9 essential traits for the carbon, water and nutrient balance of plants collected at different sites, which were selected to cover the gradient of soil and climate variation of the Amazon basin. Results showed that leaf and wood traits can be accurately estimated using a regional model based on FT NIR data. Key features such as SLA, LDMC and WD were accurately predicted by the regional model. However, the regional model did not present good results in the estimation of traits values for external sites. I conclude that FT-NIR is a promising tool for large-scale estimation of plant functional traits in the Amazon, but still a better coverage of the Amazon basin variability is necessary to produce transferable models. In the third chapter, I used the data predicted by the model generated in the second chapter to test the ability of FT-NIR predicted features to recover the pattern of functional composition across two environmental gradients, soil fertility and height above nearest drainage (HAND). I compared the patterns obtained from the data predicted by FT-NIR with the patterns obtained from conventional data, i.e., data collected directly in the field. Both datasets recovered the same pattern: the functional composition of communities tended to converge viii to what is expected by the classical cost-benefit theory of functional strategies along resource gradients: habitats with higher resource availability selected functional traits that indicate rapid growth strategies and short life cycles. The functional composition of communities did not vary strongly along the hydrologic gradient. This demonstrates that functional data predicted by FT-NIR can be used as a shortcut for large-scale ecological studies.