Assessment of soil microbial activity using VIS-NIR-SWIR-MIR spectroscopy in subtropical soils of Brazil

Extracellular enzymatic activities of soil microorganisms play a key role in driving soil nutrient cycling and exert relevant impacts on environmental, agricultural, and climatic systems. This study evaluated the potential of visible, near, shortwaves and medium infrared (VIS–NIR-SWIR-MIR, 350 to 25000 nm) spectroscopy, combined with machine learning techniques, to estimate total carbon, microbial biomass carbon (MBC), β-glucosidase, and arylsulfatase in agricultural soils under no-tillage systems in southern Brazil, including lowland and upland areas. Spectral data were acquired at laboratory, preprocessed and used to construct predictive models of soil biology based on Random Forest and partial least squares regression. The results demonstrated superior performance of MIR compared with VIS–NIR-SWIR for all evaluated soil variables, particularly for biological attributes. MIR showed high predictive accuracy for total carbon (R² = 0.94), followed by β-glucosidase (R² = 0.86), while moderate coefficients of determination were obtained for microbial biomass carbon and arylsulfatase (R² = 0.64 and 0.60, respectively). VIS–NIR-SWIR also showed good performance for total carbon (R² = 0.93) but exhibited lower predictive capacity for biological attributes.  Our results support VIS-NIR-SWIR and MIR spectroscopy as a practical tool for sustainable soil management, as it is a fast and non-destructive alternative to conventional methods that will facilitate the quantification and characterization of microbial activity.