COMPSEC, a new tool to derive natural background levels by the component separation approach: Application in two different hydrogeological contexts in northern Italy

The assessment of natural background levels (NBLs) is crucial in the management of groundwater resources and the definition of the status of groundwater bodies according to the Water Framework Directive 2000/60/EC. This work presents a new MATLAB code to assess NBLs by the component separation method, which concerns the identification and separation of a natural and an anthropogenic component within geochemical data. To increase the validity of the estimated NBL, the code calculates a set of possible solutions that should then be evaluated by the user. The following evaluation criteria are proposed: (i) stability of solution, (ii) hydrogeochemical reliability, (iii) agreement with the conceptual model and (iv) goodness of fit.The code is tested on groundwater Mn, Fe, As, Ni, Cr(VI) and NH<inf>4</inf> concentrations observed in two study areas with different hydrogeological and geochemical features in northern Italy: the multi-layer aquifer of Cremona (lower Po Plain) and the alluvial aquifer of the Aosta Plain (western Alpine sector). The pre-selection method is also applied to estimate NBLs and threshold values are assessed.The NBLs calculated through the component separation method (4292, 27 and 1220μg/L for Fe, As and NH<inf>4</inf> in Cremona and 56, 261, 9 and 2.3μg/L for Mn, Fe, Ni and Cr(VI) in Aosta, respectively) approach those estimated by the pre-selection (2798, 25 and 1820μg/L for Fe, As and NH<inf>4</inf> in Cremona and 52, 264, 10 and 1.8μg/L for Mn, Fe, Ni and Cr(VI) in Aosta, respectively). This outcome increases the validity of the estimated NBLs. The code showed versatile applicability in different contexts and conditions (e.g., different numbers of samples and monitoring points, different sampling networks). This work highlights the strength of the component separation method although some criticisms are pointed out, mainly related to the superimposition of a log-normal function for modeling the natural component.