Author(s): Renata J. Romanowicz; Marzena Osuch; Steve Wallis
Linked Author(s):
Keywords: No Keywords
Abstract: An ability to predict the fate of pollutant concentrations in river systems is necessary for assessing the ecological impact of pollution incidents and for designing appropriate river management strategies. This usually requires the application of a mathematical model that simulates the transport of the pollutant during the main processes of advection and dispersion. Since pollution incidents may occur at any flow rate, the model needs to be calibrated and validated over a wide range of flow conditions. Hence it is necessary to incorporate any flow dependence of the model parameters into the modelling system. It is usually assumed that the time-scale of changes in the flow are longer than the time-scale of any pollution incident so that all modelling takes place under steady flow conditions. This assumption may not always be true, yet very little work has been undertaken to explore the limitations of such an approach. In this paper we apply the transfer function based active mixing volume (AMV) model and the physically-based transient storage model (OTIS) to describe the transport of a pollutant in a river reach under unsteady flow conditions. For both models, tracer concentration profiles observed at two cross-sections during a number of tracer experiments performed under different steady flow conditions are used to derive the relationships between the model parameters and river flow. Appropriate optimisation methods are used to estimate the model parameters for each experiment, including information on the uncertainty of the estimated values. The derived models parameterized on flow are verified using observations not used during the calibration stage. When the models are run under unsteady flow conditions, the model parameters are updated according to changing flow using the parametric relationships with flow identified earlier. The resulting concentration profiles show that as well as having narrower confidence bounds of the predictions, the OTIS model gives about a 10% higher peak than the AMV model.
Year: 2010