Application of machine learning and multivariate approaches for source apportionment and risks of hazardous elements in the cropland soils near industrial areas in Bangladesh

Abstract

Quick industrial and urban development accelerates the hazardous elements (HEs) content in agricultural soil which is a matter of concern for ecological and public health. This present study investigates the HEs content (As, Pb, Cd, Cr, Ni, and Cu) in agricultural fields near industrial areas of Bangladesh to assess the metal concentration, distribution, sources apportionment, and their probable ecological and human health impacts using an integrated approach of machine learning and multivariate indices. Principal component analysis (PCA) showed about 28.9% of spatial variation of HEs in the study area. Metal concentrations were compared to international soil quality standards (ISQSs) and found that Cd, Ni, and Cu exceeded the ISQSs while about 80% and 100% of samples surpassed the Toxic reference value for Cd, Cr, Ni, and Cu, respectively. All HEs contents exceeded the background value excluding As. According to Positive matrix factorization (PMF) analysis, three possible sources controlled HEs content in the study area: (i) anthropogenic (Cd; 79%); (ii) mixed sources (Cu; 76%), (Ni; 51%), (Cr; 48%), and (As; 39%); and (iii) atmospheric fallout (Pb; 67%), that is corresponding by self-organizing map (SOM), and random forest (RF). The pollution load index indicated that about 77% area was polluted by studied HEs. Additionally, other pollution evolution indices including mean Effect Range Median quotient (0.31 ± 0.05), Nemerow pollution index (2.64 ± 1.13), Degree of contamination (8.20 ± 2.16), and Modified hazard quotient (1.69 ± 0.19), exhibited moderate contamination of HEs in soil samples. Based on ecological risk about 20–53% of samples were moderate to high risk, while the descending order of HEs was Cd > Pb > Ni > As > Cr > Cu. For the public health aspect, no significant non-carcinogenic health risks (THI < 1) for adults (female = 1.03E-01 and male = 9.50E-02) and children (6.83E-01), while moderate carcinogenic health risks for adults (female = 2.65E-05and male = 2.36E-05), and children (4.37E-05). Overall, this study area soils are polluted with HEs mainly by anthropogenic activities, which have significant health hazards for the inhabitants.