Assimilation of water quality buoy data for improved forecasting

Deterioration of coastal water quality poses significant threats to coastal communities. A real-time water quality monitoring and prediction system are vital in solving water quality issues for risk reduction and better management. For the Chennai coastal waters, an operational water quality forecast advisory was developed, with a dedicated web server in the public domain to track water quality on a regular basis. The goal of this work was to use a data assimilation (DA)-based model optimization strategy to improve model outputs and effectively reduce variations from real-time observation. An ensemble Kalman filter widely adopted in DA for nonlinear model dynamics has been used. Experiments with ensemble sizes ranging from 3 to 15 were conducted, and the model simulation was constrained to 10 ensemble sizes for temperature, salinity, and dissolved oxygen to evaluate the model skills. The results indicate that the DA enhances the model performance for all three water quality variables, with improved correlation resulting in a reduction of error up to 60%, generating better initial conditions, course correction, and improving model forecast skill.