A comparison of soil texture measurements using mid-infrared spectroscopy (MIRS) and laser diffraction analysis (LDA) in diverse soils

Abstract

Spectroscopic methods for the determination of soil texture are faster and cheaper than the standard methods, but how do the results compare? To address this question, laser diffraction analysis (LDA) and mid-infrared spectroscopy (MIRS) analysis have been compared to conventional sieve-pipette measurements of texture in diverse European and Kenyan soils. To our knowledge this comparison between LDA and MIRS has not been made previously. It has used soils with a broad range of organic carbon (OC) contents to investigate whether, as in other techniques, clay-OC aggregation affects the estimation of clay with MIRS. The MIRS predictions of clay content were much better than the LDA measurements, but both techniques gave good measurements of sand content. The MIRS over-estimated clay at low clay content and under-estimated at high clay content (calibration set R 2 \u2009=\u20090.83). The LDA over-estimated clay by\u2009~\u200960% (calibration set R 2 \u2009=\u20090.36), indicating that the widely used clay threshold of\u2009<\u20098\xa0µm was too high, and\u2009<\u20094\xa0µm was found to be more accurate. In samples with\u2009<\u20095% OC content, both the LDA and MIRS gave very good clay predictions (R 2 \u2009=\u20090.88 and 0.81, respectively). But in predictions of clay content in samples with\u2009>\u20095% OC the LDA under-estimated (R 2 \u2009=\u2009\u2009<\u20090.1) and MIRS over-estimated (R 2 \u2009=\u20090.34) clay content. In soils with OC removed, the MIRS prediction of clay content improved, indicating interference between over-lapping spectral regions for organic and mineral constituents. Unlike granulometric measurements of texture such as the LDA, MIRS analysis is not subject to the limitations imposed by the shape and density of particles. It was concluded that in typical agricultural soils with\u2009<\u20095% OC and\u2009<\u200960% clay content, both techniques could be used for cheap, fast and reliable estimates of soil texture.