Dados do Trabalho

Título

VIS NIR-SWIR AND MID IR SPECTRAL CHARACTERIZATION BASED ON MACHINE LEARNING TECHNIQUES OF SOIL MICROBIOLOGICAL ACTIVITY ASSOCIATED WITH ORGANIC MATTER FRACTIONS

Resumo

The microorganism’s activity is related to the degradation of soil organic matter (SOM) and involves intra and extracellular enzymatic activity. The SOM fractions are key to the understanding it and the destination of the stabilized C. The SOM fine mineral fraction (MOAM) contributes more to its conservation compared to particulate SOM (POM), while POM contains the microbial activation material, such as oxygen-containing functional groups that are preferentially utilized by microorganisms. This type of analysis can be assisted by new techniques, such as spectroscopy. This work aims to characterize soil samples by Vis-NIR-SWIR and Mid-IR spectral ranges by qualitative and machine learning techniques to access bands related to microbial activity. We analyzed the microbial biomass carbon, enzymatic activity (β-glucosidase, urease and phosphatase) and SOM fractionation. Thus, an interpretation of the spectral data of MOAM and POM was performed, supported by literature and machine learning techniques, defining the most important bands. It was observed that the spectral peaks reported for the functional groups CH, NH, COH, CO and PO are related to the activity of the enzymes analyzed. B-glucosidase for example, is related to labile compounds of the MOS so pronounced peaks were identified in the POM fraction, whose amplitude is considerably reduced when analyzed in the MOAM. This corroborates that the POM fraction includes the organic compounds, that activate this enzymatic degradation and favors the discrimination of the different enzymes, without discarding a joint analysis of the fractions that helps to identify the activity of the microorganisms. Qualitative analyses and machine learning would favor the quantification and characterization of the microbial diversity, understanding of the destiny of the C forms, C sequestration potential and consequently the evaluation of soil quality.

Palavras-chave

soil spectroscopy; soil quality; enzymatic activity

Instituição financiadora

The São Paulo Research Foundation (FAPESP) n 2021/05129-8 and Ministerio de Ciencia, Tecnología e Innovación of Colombia Scholarship Program No. 860.

Agradecimentos

The authors are grateful to the members of Geotechnologies in the Soil Science Group (GEOCIS) (https://esalqgeocis.wixsite.com/english)