Quantifying Summer Internal Phosphorus Loading in Large Lakes across the United States.

Abstract

Internal phosphorus loading (IPL) can be a significant phosphorus (P) source in freshwater systems, often causing water-quality improvement delays. Despite its importance, IPL estimates are missing for many freshwater systems due to several large-scale measuring and modeling challenges. In this study, we develop a modeling framework to estimate summer anoxic sediment release rates (SRRs) for P in 5899 large lakes and reservoirs (surface area > 1.0 km2; mixing depth < maximum depth) across the contiguous US (CONUS). Our framework combines random forest models for bottom-water temperature (BT) and surface-water total P (TP) with a mixed-effects regression model for SRR, and it includes uncertainty propagation across these models. Our results indicate that mean summer SRR ranges from 1 to 37 mg/m2/day across CONUS lakes, with 31% of waterbodies having SRR > 10 mg/m2/day. Areas of high SRR are generally associated with high predicted surface-water TP, which is particularly common in agricultural areas. Uncertainties in SRR predictions are largely attributable to the random forest-based inputs and predictive error in the SRR regression. In relatively dry summers, IPL is likely to be higher than external loading in 26% of watersheds. Overall, our results reveal where IPL can be a critical factor in watershed nutrient management.