The empirical relationship between total phosphorus and chlorophyll has guided lake management decisions for decades, but imprecision in this relationship in individual lakes limits the utility of these models. Many environmental factors that potentially affect the total phosphorus-chlorophyll relationship have been studied, but here we hypothesize that imprecision can be reduced by considering differences in the proportions of phosphorus bound to three different "compartments" in the water column: phosphorus bound in phytoplankton, phosphorus bound to suspended sediment that is not associated with phytoplankton, and dissolved phosphorus. We specify a hierarchical Bayesian network model that estimates phosphorus associated with each compartment using field measurements of chlorophyll, total suspended solids, and total phosphorus collected from reservoirs in Missouri, USA. We then demonstrate that accounting for these different compartments yields accurate predictions of total phosphorus in individual lakes. Results from this model also yield insights into the mechanisms by which lake morphometric and watershed characteristics affect observed relationships between total phosphorus and chlorophyll.
Keywords: Bayesian statistics; chlorophyll; eutrophication; phosphorus; suspended sediment.