Recognition of areas of anomalous concentration of potentially hazardous elements by means of a subcatchment-based discriminant analysis of stream sediments

Abstract

The statistical technique of discriminant analysis associated with the calculation of an information coefficient has been applied to the concentrations of 37 chemical elements for calculating the mixing of stream sediments of different origin in the Mignone river basin. Discriminant analysis has been based on sample catchment basins (SCBs), defined as the part of the drainage basin between two consecutive sampling points along the same stream branch, and on the identification of 4 different litho-geochemical groups. This approach, has been used to define the membership probability values for every sample by applying Bayes' rule and calculating posterior probability. The grade of uncertainty for each group assignment has been evaluated by using an information coefficient, based on a classification entropy index, and running a procedure analogue to that used for processing membership values in fuzzy analysis. The maximum theoretical concentration that can be expected in soils near the sampling point ( enhanced concentration) has then been calculated from both the measured and the membership values by introducing a specific enhancement function. Theoretical background concentrations at every sampling point have been also calculated by weighting the average value of concentration in each group with the membership values for each sample. These have been successively compared with the measured and enhanced concentrations to identify anomalous areas. The distribution maps of Arsenic and Vanadium in the Mignone River basin (central Italy) have been drawn accordingly to this technique, leading to the identification of areas of potential risk for human health.