AUTHOR=Nawar Said , Mohamed Elsayed Said , Essam-Eldeen Sayed Safa , Mohamed Wagih S. , Rebouh Nazih Y. , Hammam Amr A. TITLE=Estimation of key potentially toxic elements in arid agricultural soils using Vis-NIR spectroscopy with variable selection and PLSR algorithms JOURNAL=Frontiers in Environmental Science VOLUME=11 YEAR=2023 URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2023.1222871 DOI=10.3389/fenvs.2023.1222871 ISSN=2296-665X ABSTRACT=

Potentially toxic elements (PTEs) pose a significant threat to soil and the environment. Therefore, the fast quantification of PTEs is crucial for better management of contaminated sites. Versatile technique such as Visible near-infrared spectroscopy (Vis–NIRS) (350–2,500 nm) has attracted tremendous attention for assessing PTEs and has achieved promising results combined with successful multivariate analysis. This research investigated the potential of Vis–NIRS combined with partial least squares regression (PLSR) and variable selection methods to assess key PTEs (Cd, Co, Cu, Cr, Pb, and Zn) in agricultural soils under arid conditions. The soil samples (80) were collected from a polluted area around Al-Moheet drainage, Minya Governorate–upper Egypt. The samples were scanned using an ASD FieldSpec-4 spectroradiometer. Simulated annealing (SA) and uninformative variable elimination (UVE) were used to select the effective wavelengths in predicting PTEs. PLSR was used to develop the spectral models using the full range (FR-PLS) and feature-selected spectra techniques SA (SA-PLS) and UVE (UVE-PLS). The results indicated that UVE-PLS models performed better than FR-PLS and SA-PLS models in predicting the key PTEs. The obtained coefficient of determination (R2) and the ratio of performance to deviation (RPD) were 0.74 and 2.48 (Cr), 0.72 and 2.03 (Pb), 0.62 and 1.86 (Cd), 0.59 and 1.78 (Cu), 0.52 and 1.68 (Co), and 0.46 and 1.41 (Zn), respectively. The results suggested that the UVE-PLS spectral model is promising for predicting Cr, Pb, and Cd, and can be improved for predicting Cu, Co, and Zn elements in agricultural soils.