Premium
Optimizing acquisition parameters in diffuse reflectance infrared Fourier transform spectroscopy of soils
Author(s) -
Deiss Leonardo,
Margenot Andrew J.,
Culman Steve W.,
Demyan M. Scott
Publication year - 2020
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.1002/saj2.20028
Subject(s) - diffuse reflectance infrared fourier transform , soil water , spectral resolution , spectroscopy , chemistry , infrared , fourier transform infrared spectroscopy , resolution (logic) , dispersion (optics) , analytical chemistry (journal) , diffuse reflection , materials science , soil science , environmental science , optics , spectral line , environmental chemistry , physics , computer science , artificial intelligence , biochemistry , photocatalysis , quantum mechanics , catalysis , astronomy
The number of co‐added scans and spectral resolution are two fundamental acquisition parameters in diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS). However, systematic studies evaluating these parameters in soil science applications are lacking, especially when using mid‐infrared frequencies (mid‐DRIFTS). The objective of this study was to evaluate how numbers of co‐added scans and spectral resolution affect both qualitative and predictive applications of high‐throughput mid‐DRIFTS of soils. We first explored the literature with a systematic review to better understand variability in acquisition parameters and their hypothesized relationships with mid‐DRIFTS predictive performance (accuracy) of soil organic and/or total carbon (as a reference variable), but no evident relationship could be established across studies. Second, we experimentally evaluated how spectral resolutions (4, 8, 16, and 32 cm −1 ), and number of co‐added scans (8, 16, and 24 scans) affected (i) specific spectral peaks representing mineral and organic functional groups, and (ii) predictive performance of soil variables clay, sand, pH, total organic carbon, and permanganate oxidizable carbon (POXC). Decreasing the number of co‐added scans from 24 to 8 increased wavenumber‐specific spectral variability and decreased both the predictive performance and the ability to characterize smaller peaks of mineral and organic functional groups. In contrast, broadening spectral resolution from 4 to 32 cm −1 reduced multivariate scores dispersion and had a positive effect on the predictive performance, even though some smaller peaks disappeared in resolutions wider than 8 cm −1 . Acquisition parameters can be set to reduce both scanning time and computational demand while maintaining qualitative and predictive applications of mid‐DRIFTS in soil analysis.