Characterising and quantifying soil clay-sized minerals using mid-infrared spectroscopy

Characterisation of clay-sized minerals in soils is critical for assessing soil quality and function. However, the lack of high-throughput methods to quantify clay-sized minerals limits the incorporation of such information into large-scale soil surveys and assessments. This study aimed to predict the quantity of clay-sized minerals and the dominant mineral types using mid-infrared (MIR) spectroscopy in an applied manner. A total of 7193 samples were retrieved from the Kellogg Soil Survey Laboratory (KSSL) dataset. Eight pedologically significant minerals, namely calcite, chlorite, gibbsite, kaolinite, mica, montmorillonite, quartz, and vermiculite, were chosen as the target minerals. The clay-sized minerals were quantified using X-ray diffraction (XRD) analysis, categorised into six grades (grade 0 indicated not present and grades 1–5 based on increasing quantity). The soil MIR spectra showed distinctive features related to their mineralogy, particularly for calcite, gibbsite, and kaolinite. Partial least squares regression (PLSR) successfully predicted the mineral content, with most minerals having root mean square errors less than one grade. Calcite, kaolinite and mica predictions were the best, attributed to their unique absorption peaks. The prediction of dominant clay-sized minerals had an accuracy of 69 %, with calcite and montmorillonite demonstrating user’s and producer’s accuracy exceeding 75 %. A comparison between the predicted dominant minerals and the top two dominant minerals identified by XRD analysis on the samples revealed an agreement in 85 % of the cases. This study demonstrated the ability of MIR spectroscopy to predict clay-sized minerals and can serve as a rapid way to determine the dominant clay-sized minerals in soils. When following the standard procedures for both methods, MIR spectroscopy can reduce the time required for analysis by over 90 % compared to XRD, representing a significant advantage for large-scale soil surveys and mapping in the future. • Eight minerals were quantified using MIR spectroscopy and PLSR models. • 7193 samples with MIR spectra and XRD measurements were used to construct models. • VIP scores revealed features of calcite, gibbsite, and kaolinite used in the models. • 85 % of MIR predicted minerals matched the top two dominant minerals from XRD.