Macro-Scale Spatial Modeling Reveals the Role of Soil Organic Matter Quality in CO2 Emissions

Abstract

[Abstract] Carbon dioxide (CO2) emissions from agricultural soils result from the interaction of various factors that regulate the quantity and quality of organic material. Our objective was to assess changes in soil CO2 fluxes following the renewal of degraded pastures using FTIR spectroscopy and a machine learning approach. Additionally, we aimed to understand the spatial distribution of organic compounds obtained from FTIR spectra bands and their correlations with soil attributes. The study was conducted in Selvíria, Mato Grosso do Sul, Brazil, in two areas dedicated to extensive beef cattle farming. Geostatistical grids were established in the study areas to evaluate CO2 emissions, and soil sampling was performed for FTIR spectral analysis and other soil attributes. The semi-quantitative analysis of the CH vibrational region involved spectral deconvolution guided by the positions of the peaks obtained from the second derivative and Gaussian curve fitting. Our results indicate that the relative abundance of compounds related to peaks at 2853 cm−1 and 2923 cm−1 showed linear and spatial correlations with CO2 emissions. Pasture renewal, followed by sorghum cultivation intercropped with Urochloa brizantha, resulted in increased soil pH up to 5.6 and reduced concentrations of compounds related to the aliphatic CH band. This management practice also led to a decrease in CO2 emissions to a range of 0.94 to 0.97 μmol m−2 s−1. The CH2 index, calculated as the ratio of peaks at 2853 cm−1 and 2923 cm−1, exhibited similar changes and could serve as an indicator of soil organic matter loss and CO2 emissions in these systems. Our findings suggest that FTIR organic compounds are spatially dependent and coupled with a machine learning approach is sensitive to distinguish chemical changes in soil organic groups and predict soil CO2 emissions.