Models to predict water chemical cluster variables

This study is an attempt to quantify and rank variables of significance to predict mean values of lake pH and related variables (alkalinity, conductivity, hardness, etc.) in small glacial lakes. The work is based on a new and extensive set of data from 95 Swedish lakes and their catchment areas. Several empirical models based on catchment and lake morphometric parameters have been presented. These empirical models can only be used to predict mean values of these variables for lakes of the same type, and these models based on geological map parameters can evidently not be used for highly time-dependent and site-typical predictions. Various hypotheses concerning the factors regulating the mean values of the cluster variables were formulated and tested. Different statistical tests were used to separate random influences from causal. The most important map parameters were: the percent of rocks and open (=cultivated) land in the so-called near area to the lake as determined with the drainage area zonation (DAZ) method], mean depth, linked to resuspension and the form and size of lakes, relief of the drainage area and lake area. Each of these variables only provides a limited degree of (statistical) explanation of the variability in mean annual values of pH and the water chemical cluster variables among the lakes. The predictability of some of the models can be markedly improved by accounting for the distribution of the characteristics in the drainage area. The variability in mean annual values of pH (and related variables) from other parameters, such as specific anthropogenic load, etc., may then be quantitatively differentiated from the impact of these geological parameters. This paper also gives a simple method to estimate natural, preindustrial reference values of these water chemical variables from the presented models.