Ensemble data assimilation methods for improving river water quality forecasting accuracy

Area of Science:

• Environmental science
• Water resource management
• Computational modeling

Background:

• River water quality is a significant global challenge.
• Real-time water quality prediction is vital for managing issues like pollution and harmful algal blooms.
• The Yeongsan River in South Korea faces water quality challenges.

Purpose of the Study:

• To enhance the accuracy and skill of water quality forecasts.
• To investigate the effectiveness of three ensemble data assimilation methods.
• To compare the performance of traditional Ensemble Kalman Filter (EnKF) and two related algorithms (Dud-EnKF and EnKF-GS).

Main Methods:

• Utilized three ensemble data assimilation (DA) methods: EnKF, Dud-EnKF, and EnKF-GS.
• Conducted twin experiments with synthetic data for algae and phosphate concentrations.
• Performed real-world experiments assimilating actual observational data.

Main Results:

• All three DA methods improved water quality forecast accuracy and skill in twin experiments.
• Downstream model accuracy improved with similar performance across methods.
• Upstream accuracy was slightly reduced due to spurious correlations; ensemble size did not significantly impact results.
• Real-world experiments showed less improvement than twin experiments.

Conclusions:

• Ensemble data assimilation methods effectively improve river water quality forecasts.
• Model accuracy can be further enhanced through refined state variable definitions, perturbation strategies, and model calibration.
• The choice of DA method and ensemble configuration requires careful consideration for optimal forecasting performance.