Dados do Trabalho

Título

DETECTION OF NITROGEN AND HUMIC COMPOUNDS IN FERRALSOLS BY PROXIMAL SENSING

Resumo

Detailed knowledge of soil constitution is relevant to determine proper management. Thus, remote sensors allow a quick and efficient prediction of their properties. In this work, the objective was to predict the contents of ammonium (NH4+), nitrate (NO3-), and humic compounds of organic matter called fulvic, humic, and humine acids, using the wavelengths of the Visible and Near Infrared spectrum – (Vis-NIR). Eighty soil samples were determined (0-20 and 80-100 cm depth). We used the FieldSpec hyperspectral spectrometer (350-2500 nm) in the laboratory. Descriptive analyzes were performed on the spectral signatures. The content of each element was determined by chemical analysis. Regression methods such as Cubist, Partial Least Squares Regression (PLSR), Random Forest, and Support Vector Machine (SVM) were applied to the data to generate the models. The SVM model achieved the prediction for, ammonium (R2 = 0,38), nitrate (R2 = 0,60), total nitrogen (R2 = 0,52), total organic carbon (R2 = 0,97), fulvic acid (R2 = 0,49) and humine (R2 = 0,49). The results showed that the wavelengths sensitive to each element are, ammonium (2115, 2120 nm), nitrate (760, 1442, 2088 nm), total nitrogen (2250, 2330, 2430 nm), total organic carbon (410, 550, 1870 nm), fulvic acid (400, 700 nm), humine (460, 570 nm). The use of Vis-Nir spectroscopy for the detection of nitrogen and humic compounds in medium sandy soil proved to be a viable, economical, and non-destructive alternative to chemical laboratory methods. In future studies, we plan to apply the model to new databases, improving the predictive capacity.

Palavras-chave

Vis-Nir; nitrate; ammonium; soil health; spectral interpretation.

Instituição financiadora

Coordination for the Improvement of Higher Education Personnel - Brazil (CAPES) Code 001

Agradecimentos

Soil Science Geotechnologies Group – GeoCiS, and UNOESTE Soils Group.