Quantification of glacial till chemical composition by reflectance spectroscopy

Chemometric modelling of soil element concentrations from diffuse visible and near-infrared (VSWIR, 350-2500 nm) reflectance spectroscopic measurements holds potential for soil element analyses. Research has demonstrated it particularly for organic agricultural soils, yet little is known about the VSWIR response of glacial tills. Soils with low organic matter content developed on unstratified glacial materials were studied at two geologically similar sites on the mafic metavolcanic rocks of the Lapland Greenstone belt in northern Finland. The till samples (n = 217) were composed primarily of quartz, plagioclase and amphibole having less than 3% of clinochlore, talc and illite. VSWIR spectra of mineral powder (<0.06 mm) samples were measured in the laboratory, and the spectral reflectance was regressed against partial elemental concentrations of till obtained by inductively coupled plasma atomic emission spectrometry (ICP-AES) following hot aqua regia digestion. Partial least-squares regression (PLSR) analyses resulted in simultaneous prediction (R-2 = 0.80-0.89) of several soil chemical elements such as Al (validation RMSE 1802 mg kg (1)), Ba (5.85 mg kg (1)), Co (0.86 mg kg (1)), Cr (6.94 mg kg (1)), Cu (2.54 mg kg (1)), Fe (2088 mg kg (1)), Mg (449.6 mg kg (1)), Mn (0.82 mg kg (1)), Ni (3.24 mg kg (1)), V (4.88 mg kg (1)), and Zn (0.80 mg kg (1)). The electronic and vibrational molecular processes causing absorption might be responsible for accurate predictions of major elements such as Al, Fe and Mg. However, the concentrations of other major and trace elements could be predicted by the PLSR because they were cross-correlated to spectrally active soil elements or extraneous soil properties. Therefore, the applicability of the results is highly sample set specific. Further, the results show that in local scale studies at geologically fairly homogenous areas the limited spread of the data may restrict the use of the spectroscopic-chemometric approach. This paper demonstrates the capability of laboratory VSWIR spectroscopy for determining element concentrations of glacial tills. Further work should focus on overcoming the issues of sampling scale and understanding the causality for cross-correlation in quantification of the elements. (C) 2011 Elsevier Ltd. All rights reserved.