Assessing the soil baseline values of a geologically complex territory: the case study of Basilicata region.

The middle and lower reaches of the Basento river and the whole basin of the Cavone river, in Basilicata region (Italy), underwent to a geochemical prospecting involving soil and stream sediments. Specifically, 190 topsoils were collected within a depth range between 10-15 cm from the ground level and 10 bottom soils were sampled within the depth range between 80-100 cm. 

Samples were analysed at the Life Analytics laboratory (Battipaglia, Italy), by ICP-MS following an aqua regia digestion, to determine the concentrations of 16 potentially toxic inorganic substances (As, Be, Cd, Co, Cr, CrIV, Cu, Hg, Ni, Pb, Sb, Se, Tl, V, Zn, SO₄).  

The purpose of the study has been the definition of the upper background limits (UBLs) for the investigation area. An exploratory data analysis (EDA) was conducted on the dataset to outline the main data structural characteristics. Due to the huge number of samples below detection limits (BDL) for Cd, CrIV, Hg, Se, Tl, only As, Be, Co, Cr, Ni, Pb, Cu, V, Zn and SO₄ were considered for the UBLs definition. 

The estimate of the above-mentioned values has been conducted following a series of rigorous statistical tests in line with the "Guidelines for the determination of background values for soils and groundwater" released by the Italian National System for Environmental Protection (SNPA).  

In detail, after imputing the few BDL values found in the selected variables by means of K-nearest neighbors (k-NN) algorithm, topsoil and bottom soils data were considered as a whole. Indeed, the dataset was subdivided into 7 subsets according the geopedological units identified based on the local pedological and geological features. 

The BoxCox algorithm was applied to the single subsets to normalize data distribution before any statistical treatment. Outliers were identified by mean of the Dixon’s or Rosner’s Outlier tests depending on the sample size, the observation of boxplots and Q-Q plots and the spatial location of some samples considered as hotspots.  

For each variable and for each subset, two statical indices (i.e.:  95th upper tolerance limit with 95% coverage (95UTL95) and the 95th upper prediction limit (UPL95)) were calculated. The more conservative among them was chosen as representative for the UBLs.

Results showed that the UBLs found are much lower than the guideline values set by the Italian Environmental Law (Legislative Decree 152/2006). Our findings emphasized how the use of guideline values established at a national level is often inadequate to administrate a geologically and pedologically complex territory such as Italy, favoring the chance of running into a wrong identification of local environmental hazards.