Distribution of particulate and reactive mercury in surface waters of Swedish forest lakes — an empirically based predictive model

The partition coefficient (Kd) expresses the ratio between particulate bound and dissolved metals. In lake dynamic modelling, it is used to distribute the total load between the dissolved and particulate routes of lake transport. An empirically based model for prediction and evaluation of a modified partition coefficient for mercury (Hg) in lake water has been developed using field data from 25 Swedish lakes. As a substitute for dissolved Hg determined by filtration, this study uses the easy reducible (or reactive) proportion of Hg in water samples and the modified partition coefficient is here denoted Kd*. The model relates Kd* for Hg to the best correlating of several lake specific environmental variables, resulting in an r2-value of 0.69. Kd* was found to decrease with increasing suspended particulate matter (SPM) concentration, iron concentration (Fe) and ratio of lake to drainage area (a/ADA). Interpretations of the model parameters were discussed in light of the clusters of statistically and functionally related variables they represent. The negative correlation between Kd* and SPM is possibly due to the particle concentration effect, which seems to be valid also for the definition used here for the partition coefficient. For Kd the particle concentration effect is often explained by an exponential relationship between colloids and SPM which results in relatively more colloid bound metals at low SPM concentrations. The fraction of colloid bound metals is counted with the ‘dissolved’ fraction which decreases the Kd. Influences of spurious correlations on the relationship between Kd* and SPM have been assessed and found to contribute to the statistical correlation. The negative correlation with iron can also be accounted for by Hg association with small particles (iron/manganese oxides and hydroxides, possibly in association with humic matter). The influence of the morphometric factor a/ADA is probably due to differences in Kd* with different Hg transport paths (direct deposition on the lake surface is believed to decrease Kd*). The model was tested using stability tests, which indicated that other related variables could replace the given model parameters. The model parameter interpretations, in terms of clusters of related variables would, however, have been similar.