A long-term study of temporal hydrochemical data in a shallow lake using multivariate statistical techniques

The multivariate statistical techniques R- and Q-mode factor analysis and discriminant analysis were applied to a dataset containing physicochemical parameters in lake water samples collected from a shallow (similar to 4 m) lake located in northwest Greece during a 10-years (1981-1991) monitoring program. Multivariate statistical methods were applied to separate datasets of Pamvotis Lake in order to identify the environmental factors associated with the physicochemical temporal variability. R-mode factor analysis was applied on raw and log-transformed datasets and has allowed the identification of a reduced number of factors with a hydrochemical meaning. R-mode factor analysis has shown that the temporal variability in the Lake Pamvotis depends on six main factors associated to evaporoconcentration processes, natural inflows, anthropogenic influences and phytoplaktonic biomass. The sinusoidal shape of the six factors scores plots has shown that the temporal variations caused by natural and human factors are linked to seasonality. Q-mode factor analysis and discriminant analysis were sequentially applied to (i) the log-transformed full data set of Lake Pamvotis and (ii) a subset contains variables (D.O., SAT, NO3-, NO2-, NH4+ and T.P.) which are related to the water quality of lake water samples. In both cases, three clusters of water samples were identified representing (i) a clear seasonal flocculation of the lake waters and (ii) a clustering of the water samples on the basis of water quality showing a seasonal eutrophication of the lake waters. Discriminant analysis showed four parameters (T.P., Na+, Mg2+ and Ca2+) affording more than 86% right assignations in temporal analysis of full dataset and one parameter (NO3-) to afford 96% right assignations of the subset contained water quality parameters. The application of Q-mode analysis and discriminant analysis has achieved a meaningful classification of lake water samples based on seasonal criteria. Overall, the results of this study present effectiveness of multivariate statistical techniques for evaluation of temporal changes in governing hydrochemical processes and water quality of lake waters. (C) 2005 Elsevier B.V. All rights reserved.