Building a quality index for soils impacted by proximity to an industrial complex using statistical and data-mining methods

Abstract

Differences in soil quality indices, in terms of the inherent properties of loess parent material and the potential for Zn, Pb and Cd contamination from the emissions of an adjacent large industrial complex, were determined. A set of independent variables was established for the soils, using principal component analysis (PCA) and a random forest (RF) method. The quality indices of 140 topsoil samples from the environs of an industrial complex were compared with those of reference soils taken from around the borders of the study area. Potential driving factors for the Zn, Pb and Cd concentrations were dehydrogenase activity (DHA), urease (Ure) activity and invertase activity (IA), C, N, K2O, MgO, P2O5 and soil clay content. Maps were generated to show the spatial distributions of the Zn, Pb and Cd contamination. Enrichment factors and potential ecological risks were calculated. We established that, in general, concentrations of Zn, Pb and Cd in the topsoil decreased with increasing distance from the industrial complex, and the levels of Zn and Cd exceeded established intervention values, even in some soils on arable land. The arable land was enriched in P2O5, while the highest values for K2O and MgO were found in wasteland soils. The mean C content of all the soils was about 2%, with N (about 0.2%), C/N ratio (about 12) and pH (about 6.9) in the order: arable land<meadow<wasteland. The highest DHA and Ure activity was determined in the reference (unpolluted) soils, while much higher IA was present in the wasteland soils. PCA model focused on factors connected with different soil uses, while RF model emphasised the natural resistance of the studied soils to degradation. Our results indicate that the driving factors of the soil quality index were controlled by the inherent properties of the loess parent material, rather than soil pollution.