The assessment of soil contamination by heavy metals is of high importance due to its impact on the environment and human health. Standard high-sensitivity spectroscopic techniques for this task such as atomic absorption spectrometry (AAS) and inductively coupled plasma spectrometry (ICP-OES and ICP-MS) are effective but time-consuming and costly, mainly due to sample preparation and lab consumables, respectively. In the present study, a laser-based spectroscopic approach is proposed, laser-induced breakdown spectroscopy (LIBS), which, combined with machine learning (ML), can provide a tool for rapid assessment of soil contamination by heavy metals. A dataset comprising 523 soil samples, from the areas of Mati, Kineta, Varympompi, and Evia (Greece) after the wildfires of 2018 and 2021, was employed to train and validate various ML models. The analysis focused on Cr, Ni, Zn, and Pb concentrations, utilizing environmental and human health screening values for soil classification. Two classification schemes were employed: the first identified samples "outside the danger zone" of contamination, while the second focused on samples "inside the safe zone". The models achieved over 93% performance for Cr, Ni, and Zn in the first scheme and 97% for Pb in the second. These findings demonstrate that LIBS, coupled with ML, can provide a reliable and efficient solution for preliminary assessment of soil contamination, particularly suited for large-scale operations of environmental monitoring and remediation efforts.
Read full abstract