Distinguishing recent and fossil organic matter - A critical step in evaluation of post-mining soil development - using near infrared spectroscopy
Abstrakt
he objective of this study was to assess the potential of near infrared spectroscopy (NIRS) coupled with partial least square analysis (PLS) for quantifying recently formed and fossil organic matter in post-mining soils near Sokolov, Czech Republic. NIR data were acquired from soil samples (n = 14), where recent SOC was quantified by 14C-AMS analysis, and from artificial mixtures (n = 125) of overburden, coal and material from fermentation (Oe) layer. 12 PLS analyses were performed on spectra, each separately calibrated to total C (Ctot), recent C (Crec), coal C (Ccoal) and kerogen C (Cker).
All models produced correlation coefficient of cross-validation RCV > 0.97. When validated using the independent validation set of soils, addition of soil samples into the calibration set improved soil predictions of Ctot and Crec (obtained from cross-validation) considerably.
All four models for Ctot and Crec were classified as successful as indicated by root mean square error ranging from 0.74 to 1.11. Results showed no clear effect of grinding on predictive accuracy; thus, we conclude that particle size <2 mm is sufficient for NIRS.
Our results demonstrate the ability of NIRS-PLS to quantify fossil and recent carbon in soils.