Artificial Neural Network (ANNs) for predicting petroleum hydrocarbons from heavy metals contaminated soils around fuel stations

Petrol stations are classified as a dangerous source of pollution for the human population due to the toxicity of emissions from evaporated vehicle fuels and fuel spillages. The contaminants released in the environment are mainly complex mixtures of petroleum hydrocarbon compounds (PHCs) and heavy metals, especially lead. Lead phased out as a fuel additive by the dawn of the 21st century, but some soils near old or long-standing gas stations have been contaminated. Contamination found at these sites, affecting groundwater, drinking water, and the soil, can run deep and spread over an area that extends well beyond the site's border. The correlation between heavy metals and heavier petroleum hydrocarbons (C>12) in soils from an urban area near petrol stations has been studied, looking to predict the organic concentration through inorganic contaminants concentration values. Metals were analyzed by ICP-OES and FP-XRF (Field Portable XRF), while PHCs were analyzed by GC/FID. No linear statistical correlation has been proved between Pb, Cu, Mn, V, Zn, Sn, Fe, and PHCs. The ANNs model, instead, has been demonstrated to have the capability to determine the relationships between organic and inorganic contaminants, allowing an accurate prediction of PHCs (C>12) (R2=0,86).