The Brazilian soils from a spectral library perspective: from fundamental to applications

• Ano de defesa: 2021
• Autor(a) principal: Paiva, Ariane Francine da Silveira
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Detecção espectral; Espectroscopia; Pedometria; Pedometrics; Proximal sensing; Sensoriamento próximo; Spectral detection; Spectroscopy
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11140/tde-14022022-170432/

Soil is a fundamental natural resource for the balance of life on the planet and for economic development, however most agricultural areas in the world are not fully known, the knowledge of the physical-chemical properties of the soil and its spatial variability constitute the essence for its sustainable use, planning and adequate management, aiming at greater productivity and conservation. Our objective with this work was to set up a Brazilian Soil Spectral Library (BSSL), evaluate the variability, characterize and discriminate soils from the spectra, and study how a soil spectral library can help in the routine of traditional soil laboratories. We also seek to develop an algorithm that allows the prediction of soil attributes via spectra and the availability of this prediction algorithm on the internet. Soil samples were obtained by donation by researchers and laboratories throughout Brazil, underwent a spectroscopic determination process in the range between 350 and 2500 nm (vis-NIR-SWIR) and some of them underwent analysis in the range of 2500 to 25000 nm (MIR) and X-ray fluorescence (XRF). BSSL allows extracting and associating inherent spectral information with geographic and environmental variables. With the development of BSSL, it was possible: I) to demonstrate the potential of a spectral library for the management of tropical soils, and II) to relate the spectral reflectance of the soil with regions and states, biomes, geology, soil classes and vegetation. This study proved that spectral information can be used to characterize the soil and its variation and diversity in the Brazilian soil, in addition to having managed to assemble a database with soil patterns via spectrum and soil attribute prediction models that can be accessed by the community and thus implement its use in future soil surveys. It was also possible to identify how the use of spectral analysis can be inserted into the routine of soil analysis laboratories, showing the potential of hybrid laboratories.