Groundwater and surface water quality characterization through positive matrix factorization combined with GIS...

This study aims at testing the effectiveness of Positive Matrix Factorization in characterizing groundwater and surface water quality, in terms of identifying main hydrochemical features and processes (natural and anthropogenic) that govern them. This method is applied in a hydro-system featured by a strong interrelation between groundwater and surface water and highly impacted by agricultural activities. Therefore, a holistic approach considering groundwater together with the surface water bodies, consisting in lake, several rivers and springs, was used. Multivariate statistical analysis, in particular Factor Analysis, has been proved to be effective in elaborating and interpreting water quality data highlighting the information carried within them, but it presents some limitations: it does not consider data uncertainty and it groups variables which are correlated positively and negatively. Moreover, in some cases the resulting factors are not clearly interpretable, describing each one various overlapping features/processes. Here, Positive Matrix Factorization is applied to groundwater and surface water quality data, and the results are compared to those obtained through a Factor Analysis in terms of both factor profiles and their spatial distribution through a GIS approach. Results of isotopes analysis are used to validate PMF output and support interpretation. Positive Matrix Factorization allows to consider data uncertainty and the solution respects two positivity constraints, based on the concept of chemical mass balance, which leads to a more environmentally interpretable solution. Results show that Positive Matrix Factorization identifies five different factors reflecting main features and natural and anthropogenic processes affecting the study area: 1) surface water used for irrigation, 2) groundwater subjected to reducing processes at advanced stages, 3) groundwater subjected to reducing processes at early stages, 4) groundwater residence time and 5) the effects of the agricultural land use on both groundwater and surface water. Positive Matrix Factorization leads to a more detailed understanding of the studied system as compared to Factor Analysis which identifies only three factors with overlapping information. Based on the results of this study, Positive Matrix Factorization could be a useful technique to perform groundwater and surface water quality characterization and to reach a deeper understanding of the phenomena that govern water chemistry.