How should we measure clay?

There is an international commitment to monitor soil, as reflected in the new EU soil monitoring law on 5^th July 2023.  However, direct measurements of soil properties, such as water retention, cation exchange capacity, adsorption isotherms etc. are expensive and time-consuming. Pedotransfer functions (PTFs) utilize soil properties that are easy to measure and inexpensive as predictors of these critical soil parameters. Clay content plays a crucial role in the estimation of many PTFs.  However, it is often universally estimated by particle size, which fails to capture the diversity of clay mineral types which exhibit markedly different and diverse effects on the physico-chemical behaviors of soils^{1, 2}.  Non-clay minerals often constitute a significant portion of the clay size fractions. Furthermore, clay minerals may not disperse and instead remain in larger size fractions, further complicating the understanding of the effects of clay. The crucial question arising is: “How should we quantify clay content”? We have devised a predictive modelling framework that combines soil spectroscopy analysis, which is more widely available in soil databases worldwide, with X-ray Powder Diffraction (XRPD). This approach serves as a method for predicting the clay 'mineral' content, providing a  much more useful predictor of soil properties. The 703 soil samples from the National Soil Inventory of Scotland 2007-2009 (NSIS2)³ with both high-quality XRPD and IR spectral data are used to develop a predictive model for quantifying clay mineral content from MIR spectroscopy by correlating the spectral data to the quantitative assessment of clay minerals from XRPD (reduced using powdR package in R) using Machine Learning (ML) techniques (e.g., Cubist, Random Forest). Our current study attempts to answer two key scientific questions: 1. Can spectra data in the MIR region, combined with ML algorithms, accurately predict mineral clay concentrations generated from XRPD? 2. Which machine learning proves most effective in developing a national-scale calibration model for prediction?

 

 

References:

[1] Schmitz, R.M., Schroeder, C., Charlier, R., 2004. Chemo – mechanical interactions in clay : a correlation between clay mineralogy and Atterberg limits 26, 351–358. doi:10.1016/j.clay.2003.12.015

[2] Six, J., Conant, R.T., Paul, E.A., Paustian, K., 2002. Stabilization mechanisms of soil organic matter: Implications for C-saturation of soils. Plant Soil 241, 155–176. doi:10.1023/A:1016125726789

[3] Lilly A, Bell JS, Hudson G, Nolan AJ, Towers W. 2011. National Soil Inventory

of Scotland 2007-2009: Profile description and soil sampling protocols. (NSIS_2). Technical

Bulletin, James Hutton Institute. DOI: 10.5281/zenodo.7688040.