AWWA WQTC62507

This paper reports on the implementation of a three-dimensional hydrodynamic model to predict the fate of turbidity and particulates through a drinking water reservoir. This work offers the potential to develop turbidity reduction strategies based on the nature of flows through the reservoir enabling cost-effective improvements in aesthetic water quality. Melbourne's protected catchments supply disinfected but unfiltered water to over 3 million people. This water has a reputation as being one of the world's most pleasant and safe, but has higher sediment loads in the distribution system than filtered water. Silvan Reservoir is an essential part of Melbourne's water supply system. The average detention time in the reservoir is approximately 3 months. However, given the complex nature of the reservoir, short-circuiting is highly likely and the effects of this on turbidity levels at the outlets has been largely unknown. Two different numerical models were used to simulate water flow in the reservoir: DYRESM, a one-dimensional hydrodynamics model; and, ELCOM-CAEDYM, a three-dimensional hydrodynamics and water quality model. The main objective of the modelling was to determine the main source of the turbidity measured at the outlets so that management and operational efforts can be appropriately focused. The models have a high data requirement and their development required the installation of temperature gauging on inflow streams and high-resolution thermistor chains and meteorological sensors in the reservoir. An extensive field experiment was also undertaken to track the inflowing water paths and to validate the models. The data collected indicated that circulation in the reservoir is controlled by two things: a strong daily internal wave signal, which causes large vertical excursions of the water column followed by horizontal transport; and, the off-take depths, which set the stratification. The models have shown that the fastest inflow-outflow travel time is approximately 3 hours with a dilution of about 100 times indicating that although some mixing is occurring, a portion of the inflow water is reaching the outlet with very little to no detention. The models also show that turbidity at the outlets is dominated by the inlet turbidity rather than the in-situ generation and that this turbidity is made up of slow settling particles. The faster settling particles, which are generated during storm events, cause turbidity spikes at the inlets but do not affect the outlet turbidity levels. This implies that the strategy for harvesting water from the catchments into Silvan could be reviewed based on turbidity limits and may allow for an increase in yield. Various management scenarios were modelled with the ultimate outcome being to reduce turbidity at the outlets and optimize detention times. The results will be used to optimize operation of the reservoir and focus further investigation works. The model will be used to assess the implications of future management and operational scenarios as they arise. Includes table, figures.