On the modelling of the particulate and dissolved fractions of substances in aquatic ecosystems — sedimentological and ecological interactions

In building mass–balance models for different substances in aquatic ecosystems, it is important to differentiate between the dissolved and the particulate phases. This is of fundamental importance both for the predictive power of mass–balance models and for predictions of ecosystem effects. The partition coefficient, Kd, defined as the ratio between the particulate and the dissolved concentrations, is often used to get a mathematical representation of the distribution of a substance. In chemical contexts, where the equilibrium between the concentration in dissolved and particulate forms is studied, the Kd-concept is important and justifiable to use. Kd, however, is not a constant but a variable that depends on the biological, physical and chemical conditions in the water. Kd varies both within and among aquatic ecosystems. In this paper, we demonstrate that the Kd-concept is problematic for interpretation of field data, and should be used with great caution in dynamic mass–balance models. We show that it would be more appropriate in modelling contexts to use the particulate fraction, PF, defined as the ratio of particulate bound concentration to the total concentration. PF is a target variable in dynamic models since it directly, and not indirectly as Kd, describes and distributes the mass flows of substances. Two more arguments support the usage of PF. First, it is very difficult to get reliable data on lake characteristic Kd-values since, by definition, Kd includes SPM (suspended particulate matter), which has a large inherent within-system variability. We demonstrate that PF generally shows a lower within-system variability than Kd. Second, the formulation of Kd as a ratio may lead to statistical inferences, with spurious correlations that explain part of the relationship between Kd and environmental variables. This is a mathematical inevitability, which is generally disregarded in Kd-models. One way to overcome these problems is to use PF and not Kd, when relationships with environmental variables are studied and modelled.