Quantification of Streambed-Sediment Phosphorus Storage in the Maumee River network Using Machine Learning

The Maumee River basin is the largest watershed in the Great Lakes region and plays a critical role in managing runoff into western Lake Erie. Despite decades of implementing improved field management strategies aimed at reducing nutrient inputs, particularly phosphorus (P), the persistence of nuisance and harmful algal blooms indicates that sediment- bound P from in-stream and legacy sources may sustain conditions that foster these blooms. Dynamics of this sediment-P pool have been historically overlooked. Storage of in-channel, fine-grained (<63 um), streambed sediment and fine-grained, sediment-P at a 60-m reach length were estimated for the Maumee River network. By integrating basin and buffer water quality, soils data, and various other environmental factors with stream-length partitioning, a comprehensive selection of features were developed to estimate streambed legacy fine-grained sediment and sediment-P storage. The approach employs an optimally parameterized random forest model to quantify Fine-sediment storage, focused on detailed assessments within two representative, benchmark sub-basins where field surveys from 2019 quantified near-channel erosion and fine sediment storage. The estimated values of fine sediment storage were used to estimate sediment-P storage throughout the Maumee River network.