Stochastic streamflow modelling for reservoir planning and management

Streamflow characteristics above certain threshold values can be considered as random variables, and a stochastic model can therefore be constructed for the streamflow series. Popular models have been based on the representation of streamflow series by Markov or Poisson processes. However, the data must sometimes be manipulated in order for these models to be applicable. If a stochastic model of streamflow could be constructed which was not based on these classical approaches, it would have many advantages. Using the peak discharge-duration joint distribution, we derive the probability distribution of the associated flood volumes. The distribution of interarrival times of flood peaks is derived using cross-entropy with fractile constraints. A major advantage of this approach is that a model derived in this way would be applicable at lower thresholds than most current models. Based on the distributions derived in this way, a risk methodology is suggested to assist with real-time operation of reservoirs for flood control, hydropower and water supply. -from Authors

Stochastic streamflow modelling for reservoir planning and management Academic Article