Exploring internal phosphorus loads and management interventions through Bayesian reservoir modeling

Phosphorus is a limiting nutrient for eutrophication in many lakes and reservoirs across the world. While most management strategies aim at reducing external (watershed) phosphorus sources, internal phosphorus loading (IPL) often receives less attention, as it is challenging to measure or estimate through modeling. Here, we present a novel approach to characterize internal loading dynamics, leveraging multi-decadal monitoring data and a mass-balance model developed within a Bayesian inference framework. The model performs well (R² = 57 % and RMSE = 0.032 mg/L for total phosphorus) when applied to Jordan Lake, a segmented eutrophic reservoir in North Carolina, USA. Results highlight the dominant and increasing role of internal loading in summers (0.53 g/m²/month), contributing nearly twice the external phosphorus loading. We also explore long-term lake warming scenarios that intensify fluxes both into and out of the sediment layer, but have little effect on water-column phosphorus concentrations. Additionally, we investigate long-term phosphorus dynamics under different management interventions. Our simulations demonstrate the potential for IPL mitigation (e.g., capping, dredging) to accelerate and sustain water quality improvements, particularly when paired with external loading reductions. The modeling approach is transferable to similar waterbodies, especially where there is a need to characterize internal phosphorus fluxes based largely on water-column monitoring data. It also provides a computationally efficient framework for making long-term, probabilistic forecasts accounting for future climate, anthropogenic impacts, and the gradual accumulation (or depletion) of phosphorus in the sediment layer.