Probabilistic Simulation of Large Dam Operations for Water Supply and Flood Control

A flood operations model was built using GoldSim to help better understand and fine tune operations of large dams used for water supply and flood control of Wivenhoe and Somerset Dams in Queensland Australia. GoldSim was used to probabilistically simulate thousands of flood scenarios, which made it possible to compare different operating schemes under many possible conditions. The approach was found to be valuable in the process of understanding the capacity of the dams to mitigate floods while protecting water supplies. The studies identified shortcomings in the conventional design event approach to flood estimation. A broader range of stochastic floods provided an advantage to better assess flood mitigation performance and extreme floods, which is important for dam safety.


The key components of this framework are illustrated in the image below. Catchment hydrology was analyzed and calibrated, resulting in a wide range of potential flood events that were used as input to the model. A broad range of rainfall events at various locations within the catchment were also used. The model operates on these variable influences to simulate the flood operations of the dams, which result in various flood flows (flood hydrographs). The model was developed using a "top-down" approach by capturing the key drivers and feedback influences.

The model required a "learning" method to simulate realistic limits of foresight into flood flow scenarios that were incorporated into the decision logic for the operations of the dams in simulated real time. At the start of each flood event, the full event hydrograph is not known. As the flood event progresses, more knowledge of the flood flow hydrographs from the catchments becomes available as actual rainfall occurs. The model was able to simulate quantifiable aspects of decision making for the flood operations as the progression through simulated time occured. The image below is a screen capture of the control panel of the model, which was used to inform the user of operational settings in the system during the simulated flood events.


The simulated historical flood results were found to align well with the flood mitigation performance identified from the stochastic flood simulation results of GoldSim. The plot below illustrates the comparison of the probabilistic results compared to recorded historic flood events.

It can be challenging to understand the full scope of adverse impacts and positive benefits in modifying flood operations of a gated flood mitigation dam. The task is more challenging with operating rules that consider downstream catchment flows and with realistic understanding that every flood is different due to variability of rainfall patterns. The comprehensive approach adopted for the Wivenhoe Somerset Dam Optimisation Study was found to be very useful to provide evidence based information to stakeholders to understand the implications of alternative flood operations. Conventional design flood event hydrographs using uniform rainfall probability and temporal pattern on the entire catchment are not well suited to define the flood risk. Numerous flood hydrographs generated from stochastically derived rainfall events with variable space-time patterns of rainfall were more useful to detect potential changes in extreme flood levels. This was the only method that was able to differentiate the implications of different flood operations on the interests for dam safety.

Additional Reading

A paper describing this study was presented at the Australian National Committee on Large Dams (ANCOLD) conference in 2014:

Two other studies were conducted using the GoldSim model to support their results:


Making Better Decisions In An Uncertain World