Speaker
Summary
Introduction
Soybeans are Brazil's main agricultural export product, playing a strategic role in the country's socioeconomic development. Soybean production is located in the south, southeast, and central-west regions and has been expanding northward, forming the so-called 'soybean belt' as it advances over transition areas between Cerrado and Amazon biomes. Deforestation represents a challenge for the insertion of Brazilian products into the international market. To ensure sustainability and meet traceability requirements, it is essential to have methods capable of discriminate the producing regions and respective biomes.
Description of the Work or Project
This work proposes the application of analytical techniques for the traceability of Brazilian soybeans, with an emphasis on determining geographic origin based on their elemental profile. To this end, 60 soybean samples from four Brazilian biomes: Amazon, Cerrado, Atlantic Forest, and Pampa were analyzed by inductively coupled plasma mass spectrometry (ICP-MS) and neutron activation analysis (NAA). For ICP-MS analysis, the samples were subjected to microwave-assisted acid digestion. Aliquots of 500 mg were weighed into polytetrafluoroethylene tubes, together with nitric acid, hydrogen peroxide and deionized water. Digestion was carried out in a Milestone ETHOS UP system. The solutions obtained were analyzed in an Agilent 8900 triple quadrupole ICP-MS mass spectrometer. For NAA, aliquots of 250 mg were packaged in high-density polyethylene capsules suitable for neutron irradiation. The samples were irradiated at the IEA-R1 nuclear research reactor of the Institute for Nuclear and Energy Research, Brazilian National Commission of Nuclear Energy (IPEN/CNEN), with thermal neutron flux. The induced radioactivity was measured by high-resolution gamma spectrometry using hyperpure germanium detectors. The identification of radionuclides and quantification of mass fractions were carried out using the Quantu software package1. Thirty-two elements — As, B, Ba, Br, Ca, Cd, Ce, Co, Cr, Cs, Cu, Eu, Fe, Gd, Hg, K, La, Mn, Mo, Na, Nd, Ni, P, Pb, Rb, S, Sc, Ta, Tb and Zn — were determined in all samples. The data were subjected to univariate (NPANOVA) and multivariate (PERMANOVA) statistical analyses, in addition to the application of supervised machine learning algorithms Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN) and Multilayer Perceptron (MLP) to classify the samples according to the biome of origin.
Conclusions
The results indicated statistically significant differences between chemical profiles of the biomes, evidencing the potential for discrimination. The classification models achieved accuracy of up to 92% (RF), indicating the feasibility of using the elemental profile combined with machine learning for soybean traceability.
References
Bacchi, M. A.; Fernandes, E.A.N (2003). Quantu-design and development of a software package dedicated to k0-standardized NAA. JRNC, v. 257, n. 3, p. 577–582. DOI:10.1023/A:1025496716711
| Are you interested/eligible for the Young Session? | Yes, I am eligible and interested in participating |
|---|
