Groundwater and surface water quality characterization with positive matrix factorization

Groundwater and surface water quality characterization consists in identifying the main hydrochemical features of a system by analysing experimental datasets. This requires proper statistical techniques to be carried out. Multivariate statistical analysis, in particular Factor Analysis (FA), has been widely applied for investigating water quality data. FA presents some limitations: the uncertainty of the data is not considered, and it aims at gathering together positive and negative correlated variables. In this study Positive Matrix Factorization (PMF) has been investigated as an alternative to FA (Paatero and Tapper, 1994). PMF is a multivariate analysis aimed at source identification and apportionment, specifically designed to cope with environmental data and to manage their uncertainty. PMF has been widely applied in studies concerning air pollution. A few studies demonstrated that PMF can be successfully applied to datasets concerning different environmental matrices (e.g. lake sediments and soil) to reach a more realistic representation of the sources affecting different systems. Here, the effectiveness of PMF as a tool to perform water characterization is investigated, by applying it to water quality data and comparing its results with those obtained with the more widely used FA. The study area is a part of the Oglio River basin after the outflow from Lake Iseo. It is a system characterized by strong relationships between groundwater and surface water bodies strongly impacted by agricultural landuse. This work is a result of the project “Lake, stream and groundwater modelling to manage water quantity and quality in the system of Lake Iseo – Oglio River” supported by Fondazione Cariplo (grant number 2014-1282) carried out between 2015 and 2018