Freshwater ecosystems are experiencing unprecedented levels of anthropogenic stress, challenging resource managers to maintain high water quality in the face of warming water temperatures, increasing phytoplankton blooms, and decreasing oxygen concentrations. Near-term forecasts, with quantified uncertainty, can help managers anticipate these water quality impairments. We developed and applied an iterative near-term forecasting system (FLARE – Forecasting Lake And Reservoir Ecosystems) to a drinking water reservoir to generate 16-day forecasts of multiple water quality variables. FLARE uses daily data assimilation (ensemble Kalman filter) and a 1-D hydrodynamic model coupled to an aquatic ecosystem model (General Lake Model – Aquatic Ecosystem Dynamics) to generate initial conditions and calibrate parameters for 16-day forecasts. Every day, FLARE develops probabilistic forecasts for drinking water quality variables that include physical (water temperature), chemical (dissolved oxygen concentrations), and biological (phytoplankton) processes.
Results/Conclusions
Given the increased variability facing many freshwater ecosystems, ecological forecasting has much potential for improving preemptive management and minimizing water treatment costs. Over two years of 16-day forecasts, FLARE was most skillful at forecasting physical variables (water temperature, thermal stratification) and least accurate at forecasting the biological variables (phytoplankton), with the chemical variables (dissolved oxygen) intermediate. Despite these limitations, we found that the forecasts were able to accurately predict dissolved oxygen responses to the addition of oxygen from a hypolimnetic oxygenation system. Our ensemble-based water temperature forecasts were able to serve as physical covariates for other empirical model-based forecasts that do not use GLM-AED, including chlorophyll a and methane ebullition. By creating decision-based scenarios and forecasts of multiple water quality variables, our forecasts can provide managers with data to help them better manage water quality at the reservoir. FLARE provides an open-source forecasting framework for developing real-time forecasts, updated and calibrated using high-frequency sensors, that is both generalizable to other waterbodies and applicable for drinking water management.