Characterizing and Projecting Reservoir Phosphorus Dynamics through Bayesian Mechanistic Modeling

Phosphorus (P) is often the limiting nutrient for phytoplankton production in freshwater systems, and watershed P reduction is frequently considered an integral part of lake and reservoir management. However, there is increasing evidence that external (watershed) P loading reductions may not readily produce desired water quality outcomes due to persistent internal P loading from lake sediments. Moreover, the limited temporal scope of most lake sediment monitoring and modeling studies makes it difficult to characterize the long-term dynamics of internal P loading. In this study, a mass-balance P model was developed in a Bayesian inference framework utilizing 36 years (1983-2018) of water-column P data from a major North Carolina reservoir (Jordan Lake). Through Bayesian calibration, uncertainties in P flux parameters are constrained, and the model explains nearly 60% of the variability in historical P data. The model simulates an average sediment P release of approximately 0.3 g/m2/month with large seasonal variations (0.1 g/m2/month during Jan-Mar and 0.6 g/m2/month during Jul-Sept). Simulations indicate that internal sediment P releases are 70% of the external loading in the first decade and increase to 120% in the most recent decade. Results also confirm that changes in external loadings will have limited immediate benefits for eutrophication management, as internal loading will remain elevated for several decades into the future. At the same time, increases in lake temperature due to climate change will increase P cycling rates. Using the model, we compare several long-term scenario forecasts that explore how combined changes in external loading, internal loading, and temperature are expected to control future eutrophication trajectories.