Application of mid-infrared spectroscopy for rapid characterization of key soil properties for engineering land use

Abstract

Methods for rapid and accurate soil tests are needed for the index properties of material attributes commonly applied in civil engineering. We tested the application of mid-infrared (MIR) spectroscopy for the rapid characterization of selected key stability-related soil properties. Two sample sets, representing different soils from across Lake Victoria basin in Kenya, were used for the study: A model calibration set (n=135) was obtained following a conditioned Latin hypercube sampling, and a validation set (n=120) was obtained from independent sites using a spatially stratified random sampling strategy. Air-dried ground (<0.5mm) soil was scanned using a high-throughput screening accessory for diffuse reflectance attached to a Fourier transform infrared spectrometer. The soil properties were calibrated to smoothed first derivative MIR spectra using partial least-square regression (PLS), and screening tests were developed for various limitation classes applicable in civil works using the soft independent modeling of class analogy (SIMCA). The hold-out full cross-validation coefficient of determination (r2)≥0.8 was obtained for the liquid limit (LL), linear shrinkage (LS), coefficient of linear extensibility (COLE), air-dried moisture content, (W) and cation exchange capacity (CEC). Further independent validation gave r2≥0.73 and the ratio of prediction deviation (RPD) 4.4–2.1 for LL, LS, COLE, W, CEC, plastic limit (PL), plasticity index (PI), and volumetric shrinkage (VS). The independent validation likelihood ratios for the diagnostic screening tests were: LL>55%, 4.2; PI>30%, 2.7; LS>12%, 2.4; exchangeable sodium (eNa)>2cmol (+) kg−1, 2.3; exchangeable sodium percent (ESP)>10%, 1.8; W>8.3%, 1.6, and Activity number (A)>1.25units, 1.5. MIR can provide the rapid assessment of several soil properties that yield stability indices in material testing for engineering land use. Further studies should test the ability of MIR PLS for establishing broader calibrations across more diverse soil types and the direct correlation of MIR to material functional attributes.