A Random Forest in the Great Lakes: Stream Nutrient Concentrations Across the Transboundary Great Lakes Basin

Excess nutrient inputs from agricultural and urban sources have accelerated eutrophication increased the incidence of algal blooms in Great Lakes Basin (GLB). Lake basin management to address these threats relies on understanding key drivers pollution. Here, we use a random forest machine learning model leverage information 202 monitored streams GLB predict seasonal annual flow-weighted concentrations nitrogen phosphorus, as well ratios across GLB. Land (agricultural land) land (tile drainage wetland density) emerge two most important predictors for dissolved inorganic (DIN; NO3− + NO2−) soluble reactive phosphorus (SRP; PO43), while soil type density are more particulate P (PP). Partial dependence plots demonstrate increasing with tile decreasing density. In addition, livestock densities cover correspond higher SRP:Total Phosphorus (TP) ratios. Seasonally, highest proportions SRP occur summer fall. Higher also correlated N:P (DIN:TP) Livestock operations can contribute buildup nutrients excess manure application, subsurface provide transport pathways nutrients. Given that both SRP:TP strong harmful blooms, our study highlights importance management, restoration efforts intensively managed landscapes.