Artificial neural networks to evaluate organic and inorganic contamination in agricultural soils

The assessment of organic and inorganic contaminants in agricultural soils is a difficult challenge due to the large-scale dimensions of the areas under investigation and the great number of samples needed for analysis. On-site screening techniques, such as Field Portable X-ray Fluorescence (FPXRF) spectrometry, can be used for inorganic compounds, such as heavy metals. This method is not destructive and allows a rapid qualitative characterization, identifying hot spots from where to collect soil samples for analysis by traditional laboratory techniques. Recently, fast methods such as immuno-assays for the determination of organic compounds, such as dioxins, furans and PCBs, have been employed, but several limitations compromise their performance. The aim of the present study was to find a method able to screen contaminants in agricultural soil, using FPXRF spectrometry for metals and a statistical procedure, such as the Artificial Neural Networks technique, to estimate unknown concentrations of organic compounds based on statistical relationships between the organic and inorganic pollutants. •A relationship between organic and inorganic contaminants in polluted soil exists.•ANNs model can predict POPs from metal concentrations measured by FP-XRF tool.•FP-XRF coupled to ANNs is a valid method to screen hot spots in agricultural lands.