A Dynamic-Stochastic Approach for Modelling Advection-Dispersion Processes in Open Channels

Abstract A combined stochastic-deterministic model has been developed to describe the temporal and spatial distribution of conservative substances in open channel flows. The model consists of a finite difference approximation to the one-dimensional advection-dispersion equation embedded within a stochastic filter. The time and measurement updates of the estimated concentrations and their covariance are carried out through the use of a factored form of the covariance matrix. The resulting filtering algorithm is more computationally stable than the standard Kalman filter approach. The dynamic-stochastic model is shown to perform well when it is applied to simulated observations of salinity in an Arctic estuary. It is shown that this type of modelling approach can be used as a tool in planning field experiments.