Multivariate Statistical Analysis Of Hydrochemical And Environmental Isotope Data To Characterise The Hydrogeological And Hydrogeochemical System Of The Lake Sibayi Catchment, North-Eastern South Africa

An integrated approach involving multivariate statistical analysis combined with graphical methods (Piper trilinear diagram and δ18O-δ 2H plots), and environmental isotope analyses were successfully applied to characterise the spatial distribution of hydrogeochemical parameters and their controlling factors within the Lake Sibayi catchment located in north-eastern South Africa. Bivariate and Multivariate statistical analyses (Factor and Hierarchical Cluster Analyses) were performed on 12 physiochemical parameters (variables) including pH, EC, TDS and major ions of 46 samples collected from various water sources (streams, Lake, shallow and deep boreholes). Bivariate Pearson’s correlation matrix of the measured variables revealed a strong positive correlation between EC and several major elements, which included Na+ , K+ , Mg2+, Ca2+ , Cland HCO3 - indicating their contribution to the salinity. These major ions were also found to be strongly correlated to one another with all correlations found to be significant. Factor analyses in the form of Principal components analyses were performed with the main aim of identifying the underlying factors or processes responsible for the observed hydrochemistry in the study area. The results revealed three principle factors explained about 95% of the hydrochemical variation in the study area. Most of the variance is contained within Factor-1 (69.5%), which has a high positive loading factor associated with EC, TDS, Na+ , Mg+ and Clconcentrations, interpreted in terms of the contribution of these major ions to the salinity of the water (EC and TDS). Factor-2 represents 17.7% of the total variation in the hydrochemistry and has high positive loadings for pH, HCO3 - , K+ , Ca+ and Fe concentrations. The HCO3 - , K+ and Ca+ could be the result of weathering and dissolution of carbonate minerals in calciferous Uloa and Umkwelane Formations and redox processes. The Fe concentrations could be related to leaching of ferricrete layers known to exist in the area and a result of anoxic condition within the aquifer. The variables NO3 - and SO4 - contribute most strongly to Factor-3, which explains 8.01% of the total variance. The loading for NO3 - was positive and could result from anthropogenic pollution of the shallow aquifer and streams, while that of SO4 - being negative and could result from historical marine influences. Hierarchical cluster analysis of hydrochemical data performed using the Ward method with squared Euclidean distance, grouped the water samples into two clusters, representing unique hydrochemical systems, i.e. surface water and groundwater. Each of these two clusters was in turn divided into two sub-clusters, representing stream and lake samples, and shallow and deep aquifers, respectively. These groupings were further supported by characteristic water types; namely, a Na-Cl-HCO3 facies for the river, lake and surrounding boreholes; a Na-Cl hydrochemical facies for shallow boreholes, while deep borehole samples were Na-Ca-Cl to Na-Cl-HCO3 in composition. These clustering were supported by isotopic signals that show a clear distinction between groundwater and lake water samples.