Changing climate and land use are altering oxygen regimes in lakes and reservoirs around the world. This presents an important concern for lake and reservoir management, as low dissolved oxygen concentrations can restrict habitat for aerobic organisms and severely degrade water quality. Furthermore, these rapid global changes make historical averages less informative for predicting future oxygen conditions. Forecasts of oxygen concentrations with quantified uncertainty could simultaneously help understand the ecological processes underlying changes in dissolved oxygen and allow lake and reservoir managers to better anticipate future water quality concerns.
In this study, we developed a forecasting system to predict bottom-water oxygen concentrations in Falling Creek Reservoir (FCR; Virginia, USA) 14 days in the future. Twice-weekly depth profiles of temperature and oxygen provide the only data inputs to the system. The forecasting system partitions uncertainty among initial conditions, drivers, and process components, and iteratively assimilates new observations. We used data from 2013-2016 for training, then tested the predictive ability of four alternative forecast models (temperature only, oxygen only, both temperature and oxygen, and a null persistence model) in 2018 and 2019 to determine the relative importance of temperature and oxygen as controls on bottom-water oxygen demand.
Results/Conclusions
The forecasting system successfully recreated observed oxygen dynamics over six years, and forecasts can be visualized in an interactive web interface. Forecast performance varies by year, with root mean squared error (RMSE) for the 7-day forecast ranging from 0.5–1.2 mg/L and the 14 day forecast ranging from 0.7–1.6 mg/L during the test period (2018-2019). In all years, the full forecast performs better than the null persistence model. (14 day: 0.3–1.5 mg/L RMSE improvement). The temperature-only forecast performed better than the oxygen-only forecast (14 day: 0.1–1.0 mg/L RMSE improvement), indicating that patterns of oxygen are influenced more by temperature than the baseline concentration of oxygen in the water column. These results demonstrate that bottom-water oxygen concentrations can be accurately forecasted and that near-term forecasts can be used to assess the predictive power of ecological drivers. As oxygen conditions continue to change, ecological forecasts offer new opportunities for preventative management to protect lakes and reservoirs around the world.