Diagnostic screening of urban soil contaminants using diffuse reflectance spectroscopy

Abstract

There is increasing demand for cheap and rapid screening tests for soil contaminants in environmental consultancies. Diffuse reflectance spectroscopy (DRS) in the visible-near infrared (vis-NIR) and mid infrared (MIR) has the potential to meet this demand. The aims of this paper were to develop diagnostic screening tests for heavy metals and polycyclic aromatic hydrocarbons (PAH) in soil using vis-NIR and MIR DRS. Cadmium, copper, lead, and zinc were analysed, as were total PAH and benzo[a]pyrene. An ordinal logistic regression technique was used for the screening and predictions of either contaminated or uncontaminated soil at different thresholds. We calculated the rates of false positive and false negative predictions and derived Receiver Operating Characteristic curves to explore how the choice of a threshold affects their proportion. Zinc and copper had the best prediction accuracies of the heavy metals, with 89% and 85%, respectively. Cadmium and lead had the lowest prediction accuracies, with 68% and 67%, respectively. PAH predictions averaged 78.9%. With an average prediction accuracy of 79.9%, MIR analysis was only slightly more accurate than vis-NIR analysis, which had an average prediction accuracy of 77.5%. However, vis-NIR may be used in situ, thereby reducing cost and time of analysis and providing diagnosis in ‘real-time’. DRS in the vis-NIR can substantially decrease both the time and cost associated with screening for soil contaminants.

The use of diffuse reflectance mid-infrared spectroscopy for the prediction of the concentration of chemical elements estimated by X-ray fluorescence in agricultural and grazing European soils

“…The majority of previous work aimed at developing predictive models of elemental concentrations in soils using MIR has been on this group of elements (especially Cu, Ni and Zn) with similar results to those obtained in the present study (Siebielec et al, 2004;Bray et al, 2009). In addition, successful calibrations have been reported by Reeves and Smith (2009) for Ce (RPD = 1.5), Co (RPD = 2.4), Cr (RPD = 3.0), Ga (RPD = 1.7) and Ni (RPD = 3.9), and by Bertrand et al (2002) for Cu, Mn and Zn (R 2 of 0.82, 0.80 and 0.88, respectively).…”

“…Janik et al (1998) and Minasny et al (2009) reported similar models for P determined by XRF (R 2 values of 0.60 and 0.48, respectively). Most of the previous work on developing IR models for the prediction of the metals of this group has been on Pb together with Cu (Siebielec et al, 2004;Wu et al, 2005;Wu et al, 2007;Bray et al, 2009;Moros et al, 2009). In general, the higher the correlation with Fe, the better the models developed, for example better results were obtained for Cu than for Pb.…”

“…Visible (VIS)-NIR and MIR spectroscopy, or spectral combinations of these, have been used previously to develop prediction models for elemental concentrations in soil (Janik et al, 1995(Janik et al, , 1998Bertrand et al, 2002;Siebielec et al, 2004;Wu et al, 2007;Bray et al, 2009;Reeves and Smith, 2009). Using MIR, Janik et al (1998) predicted concentrations of major elements in surface soils determined by XRF for Al, Cu, Fe, K, Mg, P, Si and Ti oxides (n = 298) with coefficients of determination (R 2 ) values ranging from 0.60 (P 2 O 5 ) to 0.97 (SiO 2 ).…”

Synergistic Use of Vis‐NIR, MIR, and XRF Spectroscopy for the Determination of Soil Geochemistry

“…The goodness of fit for the conventional pXrF calibration data set are given for comparison with and animal health (essential trace elements Co, Cu, Mn, and Zn) in a dataset of non-polluted soils. To give this result perspective, case studies in polluted environments have rarely been able to successfully deliver a full assessment of soil contamination by using either vis-NIR or pXRF methods (Bray et al, 2009;Hu et al, 2014;Kemper and Sommer, 2002;Kleinebecker et al, 2013;Malley and Williams, 1997;Siebielec et al, 2004;Weindorf et al, 2012;Wu et al, 2007;2005). In the case of pXRF, results are influenced by the digestion procedure before ICP analysis (as discussed above in Vis-NIR, MIR and pXRF prediction of soil properties section).…”

“…Assessment of a full suite of soil geochemistry from large geographical datasets has shown a high number of elements to be determined by MIR spectroscopy Al, B, Be, Bi, Ca, Ce, Co, Cr, Cu, Cs, Fe, Ga, In, K, Li, Mg, Mn, Na, Nb, Ni, P, Rb, S, Sc, Sr, Ti, Th, V, Y, Zn, and Zr (Soriano-Disla et al, 2013a;Reeves and Smith, 2009). A smaller number of studies have investigated MIR for heavy metal contamination and have conversely reported MIR to be only slightly (Bray et al, 2009) or markedly (Siebielec et al, 2004) more accurate than NIR for measuring soil metal content.…”