Evaluation of multivariate statistical methods for the identification of groundwater facies, in a multilayered coastal aquifer

Abstract

The multilayered coastal aquifer system of the southern Rhodope province in North Greece was selected to examine the applicability of multivariate statistical methods in the investigation of groundwater variability and to evaluate their performance in the presence of overlapping processes. The aquifer system which forms the main source of water supply to numerous productive wells in the area, drilled at various depth horizons, faces serious degradation of groundwater quality. A widespread sampling survey was conducted at the end of the wet season 2010 and 16 physicochemical parameters were determined for every sample collected. Multivariate statistical techniques (cluster and factor analysis) were applied twice, first including all the sampling points, and second after the removal of deep boreholes samples. The results showed that the above methods were able to separate a mixture of processes in the aquifer system (saltwater intrusion, agricultural runoff, mineral dissolution, ion exchange, connate water influence), and spatially distribute them in the study area. They also highlighted the need for knowledge of technical information regarding the boreholes and of sound hydrogeological data in the area, since the inclusion in the analysis of deep connate groundwater masses caused masking effects.