Modeling the Pb Distribution Using Support Vector Machines in Surface Soil of the Lands Surrounding the Dezful-Ahvaz Road

Investigation of heavy metal distributions in environment is vital for the remediation of contaminated soils and environmental monitoring. In this study, the Pb-concentration variation in surface soils surrounding the Dezful-Ahvaz highway and the soil parameters influencing it were investigated. For this purpose, soil sampling was performed from depth of 0-10 cm of adjacent lines with the highway by collecting three identical samples at intervals of 15, 40, and 100 m from the road side (a total of 135 samples). Then, the total Pb-concentration, organic matter, calcium carbonate equivalent (CCE), pH, electrical conductivity (EC), and soil particle size distribution (clay, silt, sand, fine sand, and very fine) were measured in each soil sample. Modeling analyses were performed using support vector machine (SVM) and multiple linear regression (MLR) methods. To investigate the model performances, some statistical indicators including the correlation coefficient (r), root mean square error (RMSE), and model efficiency factor (MEF) were calculated between the measured and the predicted values. The variable selection results using the SVMs method indicated that the CCE was the most effective factor on Pb- adsorption by soil particles at the distance of 0-15 m from the road side and the clay % and pH (followed by sand %) parameters had the highest importance coefficients at the distances of 15-45 [H1] and 45-100 m, respectively. The r coefficient and MEF values for estimation of Pb using the SVM model were 0.86 and 0.70 %, while they were 0.62 and 0.38 % for the MLR models. Therefore, according to the obtained results in this study, it appears that it would be possible to use support vector machines for determining the factors influencing the Pb- adsorption by soil surrounding the road sides and for modelling heavy metal concentration variations in soils.