Reservoir turbidity modelling using artificial neural networks and the estimation of performance indicators

• Main Author: Chan-Yan, Deborah A.
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/10379

Ensuring proper suspended sediment and turbidity control requires an understanding of water body response to turbidity-causing events. This thesis describes an approach which evaluates turbidity control using risk-based performance indicators, and which is suitable for application in a wide range of water resource problems. In particular, the performance of a water supply reservoir under major drawdown conditions is evaluated in terms of meeting water quality turbidity objectives. The case study presented is based on the Capilano Reservoir located in North Vancouver, British Columbia. Response to turbidity-causing events is simulated using artificial neural networks, and turbidity levels are estimated for both the annual and wet season periods. In addition, artificial neural network models are developed for short-term forecasting of turbidity levels at the reservoir outlet. Reservoir reliability and resilience are then estimated using FORM. Reliability and resilience estimates are derived for various scenarios of interest and, as a result, can be used to evaluate tradeoffs between meeting proposed drawdown objectives and the elevated turbidity levels that may occur within the water body. The reservoir is shown to exhibit poorer performance during the wet season, particularly in the October to December period, compared to its overall performance on an annual basis.