Methods for regional calibration - a case study using the TopNet hydrological model for the Bay of Plenty region, New Zealand

Abstract

ABSTRACT The regional parameters can be derived from physiographic factors such as topography, soil type, and vegetation. In this study the suitability of five regional parameterisation methods was investigated and assessed to provide guidance on how to select optimal parameter sets. The overarching goal was to improve the accuracy of a hydrological model when applied at the regional scale. The regional parameterisation techniques assessed were: 1) averaging of optimal parameter sets of gauged catchments, 2) calibrating several catchments with a regional performance efficiency, 3) weighted average performance criterion for calibration, 4) largest number of positive Nash-Sutcliffe coefficients and 5) a Bayesian parameter selection method. The parameterisation methods were assessed by applying TopNet in sixteen gauged catchments in the Bay of Plenty region on the North Island of New Zealand, to then predict flows for five test catchments. It was shown that any of the five regionalisation methods considered improved flow predictions of Mean Annual Low Flow when compared to applying the same model with a prescribed spatially distributed set of the key model parameter. The improvement was less pronounced for mean flow and high flow, regardless of the method used and was linked to the hydrogeology class of the catchments.