Assessment of a mathematical model for runoff prediction in New Zealand

Abstract

A major problem involving the planning of water resource development in New Zealand is the scarcity of streamflow data. Long-term rainfall data is available for most regions. A mathematical model may be fitted to a catchment by minimizing the difference between the recorded and simulated runoff i.e., minimizing the response function. The model may then be used to generate runoff from historical rainfall input. The factors determining the type of model which may be used to extend streamflow records were reviewed. The Boughton model was modified to suit New Zealand conditions. The model was tested on six catchments and shown to give satisfactory results. The use of the model on ungauged catchments requires that the relationship between model parameters and catchment characteristics is determined. This problem is complicated by interdependency between the parameters which makes it difficult to determine stable optimum values. The stability of parameters was studied and it was found that the model may be fitted to a catchment to give satisfactory results by optimizing a small number of independent parameters. However, because the response function contains more than one minimum, the optimization method converged on the local minimum as determined from different starting values of the parameters. This makes it difficult to determine the absolute optimum. The physical significance of the parameters was discussed and several improvements to the model were suggested.