Estimation of the bioaccessible arsenic fraction in soils using near infrared spectroscopy

Abstract

A self-modelling mixture algorithm was used to convert raw NIR spectra of soils, developed over Jurassic ironstones, into five underlying spectral components (SC) and associated coefficients. The five SCs were shown to be significantly correlated to the total As, bioaccessible As and total Fe contents of the soils and tentatively assigned to crystalline Fe oxides, Fe oxyhydroxides and clay components in the soils. A linear regression model, using the SC coefficients associated with the clay fraction, the Fe oxyhydroxides and the total As content of the soils as independent variables, was shown to predict the bioaccessible As content of the soils (as measured by an in vitro laboratory test) with a 95% confidence limit of ± 1.8 mg kg⁻¹ and a median R² value of 0.80. The approach could be extended to other soil types but the model presented here only applies to the Jurassic ironstone soils from Lincolnshire UK.

Keywords:

• Bioaccessibility
• arsenic
• soil
• regression
• NIR
• mixture resolution
• chemometric
• Jurassic ironstone

Acknowledgements

The authors would like to thank Bruker instruments for the loan of the of the NIR instrumentation. This paper is published with the permission of Professor John Ludden, Executive Director of the British Geological Survey.

Notes

**Sample 29 was identified as an outlier and subsequently removed from the regression analysis.

*With one outlier removed (As concentration of 295 mg kg^{− 1}).