Rapid assessment of regional soil arsenic pollution risk via diffuse reflectance spectroscopy

- Soil contamination with As is a serious environmental concern.
- We used DRS to predict soil As content of 200 soils collected from a landfill site.
- ENET model using VisNIR spectra produced best As prediction.
- DRS prediction followed by kriging produced spatial variability map of soil As.
- DRS prediction with indicator kriging identified As pollution hotspots.

Soil arsenic (As) contamination by anthropogenic and industrial activities is a problem of global concern. This pilot study demonstrates the feasibility of adapting the diffuse reflectance spectroscopy (DRS) approach using the visible near infrared (VisNIR) spectra for detecting soil As pollution. Further, spatial variability of soil As contamination was evaluated combining DRS based predictions and two geostatistical algorithms. The raw reflectance spectra were preprocessed using three spectral transformations for predicting soil As contamination using three multivariate algorithms. Quantitatively, better accuracy was produced by the elastic net-first derivative model (R2 = 0.97, residual prediction deviation = 6.32, RPIQ = 7.33, RMSE = 0.24 mg kg− 1). The prediction of soil As was dependent on the close association between soil As and spectrally active soil organic matter and Fe-/Al-oxides. Moreover, the As pollution risks hotspots were reasonably identified using ordinary kriging and indicator kriging interpolations based on DRS predicted As values.