ABSTRACT
A coupled data-driven and 3-dimensional (3D) process-based fluorescent dissolved organic matter (fDOM) prediction model was developed for a shallow, subtropical Australian reservoir. The extent to which reservoir water volume, inflow, and wind conditions affect the fDOM transport dynamics during cyclonic weather events was assessed through scenario analysis and a data-driven Bayesian network (BN) approach. The analysis shows that (a) inflow plumes are the main sources of fDOM during heavy rainfall; (b) the concentration of fDOM near the dam wall is related to rainfall intensity; (c) higher reservoir volumes reduce the rate of increase and peak of fDOM concentration during rainfall events; and (d) fDOM transport to the dam wall is strongly influenced by the prevailing wind direction. A naïve BN developed for fDOM assessment displayed a strong sensitivity of the peak fDOM value to rainfall-related characteristics while the lag time between rainfall event and fDOM peak at the dam wall was highly sensitive to reservoir water volume and wind speed. The hybrid modelling approach provides both new information on 3D fDOM transport in reservoirs during extreme weather events through the model application and an easy-to-interpret, instantaneous modelling output for treatment operators through the BN modelling component. The BN modelling is an essential addition for water treatment operators to promptly predict the impacts of extreme weather events and proactively adjust treatment operations without the computational time burden of a 3D process-based model.
Acknowledgements
The authors acknowledge DHI Water and Environment, Denmark, for their assistance in providing MIKE modelling system for this study. This research work was conducted with the technical support of Griffith University and Seqwater. The authors acknowledge Jonathan Creamer, Michael Bartkow, Paul Fisher, Rohan Campbell, and David Roberts for technical advice.
Disclosure statement
No potential conflict of interest was reported by the author(s).