Part 1 of this study assessed the effectiveness of dispersion models based
around random particle tracking (RPT) applied an overbank flow in the Flood
Channel Facility at HR Wallingford. Several of the RPT models that were us
ed predicted the mixing behaviour of the tracer in the complex channel flow
to within an acceptable accuracy. If there is no uniqueness in the model s
tructure which can be used to represent the true system to within the limit
ations of the available observations, then this implies an inherent degree
of ambiguity in our knowledge of the physically based model structure. This
suggests that we should be less forthright in the optimisation of each ind
ividual model structure, and perhaps investigate more of the parameter comb
inations for each model which yield feasible simulations of the system. An
alternative fuzzy calibration technique is introduced which avoids the opti
misation process and takes account of uncertainties in the model structure,
parameter sets and observed data in prediction.