Accounting for hydrological controls when clustering groundwater quality data

Abstract

<p>Groundwater salinization occurs when high concentrations of water-soluble salts are present in groundwater systems and is regarded as one of the most worldwide, severe, and complex phenomena affecting coastal aquifers. Salinization might occur, for example, from: (i) from marine sources via seawater intrusion, seawater ingression, (ii) underground or terrestrial sources (e.g., natural soils and rocks through the dissolution of soluble minerals), and (iii) salt and saline fluids from anthropogenic activities.</p><p>The analysis of salinization-related data by multivariate statistical methods is often undertaken in the context of efficient groundwater management. For example, clustering algorithms are used to delineate hydrogeohemically distinct water classes, whereas dimensionality-reduction algorithms are being used to decipher underlying natural and anthropogenic influences responsible for these distinct water classes. However, most of these algorithms do not explicitly account for spatial information and/or constraints, which can often have a significant impact on the classification of groundwater quality samples.  In the context of the MEDSAL Project (www.medsal.net), such spatial effects are incorporated in the clustering procedure via the inclusion of pertinent hydrological attributes, namely, hydraulic head and conductivity data, along with pair-wise distances between sample locations.</p><p>The application of the proposed spatially explicit clustering approach is illustrated using groundwater quality samples collected from the Rhodope coastal aquifer, located at north-eastern Greece. Sampling locations were grouped into four hydrogeochemically distinct water classes using k-means clustering with and without accounting explicitly for spatial information. Principal component analysis (PCA) was used to decipher underlying natural and anthropogenic influences responsible for these distinct water classes. The first four principal components (PCs) explained more than 83% of the total variance in water quality variables, from which the major component was found to be associated with salinization processes.</p>