Comparison of one-dimensional and two-dimensional computer models of water flow over flood plains using the 1986 and 1994 floods on the Waihao River Flood Plain

Abstract

This dissertation compared the ability of a one-dimensional computer model and a two-dimensional computer model to model water flow over a flood plain. It used observed data from 1986 and 1994 flood events on the 50km2 Waihao River Flood Plain. The observed data was collected by a survey of the flood plain residents. The data was generally given at buildings and other easily remembered locations and also from photographs and videos of the flood event. The photographs verified the flood levels given by the residents and provided more flood levels. The Canterbury Regional Council also provided a map of the areas flooded from these floods. Further data from the Canterbury Regional Council on river discharges, stopbank overtopping and breaching was analysed to give flows onto the flood plain. The details of the Waihao River Flood Plain topography were stored in a Digital Terrain Model in a computer package called ARC/INFO. The details of the flood plain flow resistances of the flood plain using Mannings ‘n’ values were digitised into ARC/INFO. This information was imported into the computer models using software designed for this purpose. The dissertation used the Danish Hydraulic Institute MIKE11 computer model for the onedimensional analysis and for the two-dimensional analysis a computer program under development by C. Beffa of Switzerland.

The results showed that the two-dimensional model gave results with less standard error than the one-dimensional model. A calibration to obtain the best fit to the surveyed flooded area was undertaken with the two-dimensional model showed that in this case the observed flood levels were underpredicted. The results showed that for the two-dimensional model the standard error of the differences in the calculated and surveyed flood levels was less than expected error from the digital terrain model and the flood levels. This lesser standard error also validated the collection of observed flood level data from flood plain residents. The dissertation showed that both models MIKE11 and 2de could give good results, with the two-dimensional results marginally better, for the flooding in areas where recent flooding had not occurred which could have been used for calibration. The dissertation finally made recommendations to improve the ability of both one-dimensional and two-dimensional models to model water flow across flood plains.