Regionalisation of Rainfall-Runoff Models

A new approach to regionalisation for prediction of flow characteristics in ungauged catchments is presented. If flows in ungauged catchments are to be predicted using calibrated rainfall-runoff models, regional relationships between the parameters of such models and catchment attributes must be determined. This is only possible with parsimonious models (fewer than about 7 parameters), and even then there are considerable uncertainties in the prediction of ungauged catchment response, due to the accumulation of uncertainties in the regionalization process for estimating parameter values. The uncertainties in the catchment attributes and climate data used for the ungauged catchment combine with the uncertainty in the relationships between catchment attributes and model parameters for the gauged catchments. Moreover, these relationships are subject to uncertainty in both catchment attributes and gauged-catchment model parameter values. The latter are influenced by model structure, parameter identifiability, subjective selection of “optimal” parameter values and errors in the climate data. The uncertainty for the ungauged catchment can be minimised either by minimising each contributing uncertainty or by altering the approach so as to bypass some of them. An example of the second option is to regionalize catchment response characteristics rather than model parameters, then relate the ungauged-catchment model parameter values to these response characteristics. This bypasses the influence of model structure, parameter identifiability and subjective selection of “optimal” parameter values, thereby reducing uncertainty in the estimates of the response characteristics at the catchment scale. Possible response characteristics are the mean annual runoff coefficient, slope and shape of the flow duration curve, fraction of time with no flow, and the average-event flow-response curve. Initial regionalisation of these characteristics will be illustrated for a selection of catchments in Australia.