An integrated approach to inflow forecasting for reservoir management: The upper Waitaki

Abstract

In 1992 New Zealand experienced one of its worst electricity shortages in history. This was blamed on the drought of 1991/1992 and the management of the water stored in the hydro lakes. A review committee was set up by Government to consider the possible causes of those shortages and to make recommendations that could lead to avoiding a repeat of such a situation (Davison, 1992). Among the many recommendations made, the need for further research was proposed, to better understand the hydrology of the watersheds that feed the storage lakes and in predicting inflows to those lakes.

This study is aimed at addressing this recommendation by the use of an integrated approach to forecasting inflows to the storage lakes of the Upper Waitaki: Lake Pukaki, Lake Tekapo and Lake Ohau. The integrative approach involves the use of multiple discriminate analysis (MDA) for estimating annual and seasonal precipitation amounts, a multivariate chaining (MVC) process for distributing these total precipitations to daily amounts and associated minimum and maximum temperatures, and the UBC Watershed Model for simulating flows from the generated daily weather data.

The use of MDA for predicting weather is not new (Miller, 1962). In this study the predictor variables are selected from sunspot numbers, Southern Oscillation Index (SOI), Quasi Biennial Oscillation, precipitation amounts and temperatures. Possible additional predictor variables are sea surface temperatures and sea surface pressures. The MDA was found to be useful for stations in the Upper Waitaki, though not exceptional. The MVC models simulated the daily weather data very well and preserved all the main statistical characteristics for precipitation and temperatures for individual stations.

Simulations of the 1992 drought with the integrative approach was satisfactory. Retrospectively, if such an approach had been applied prior to that drought it would have been possible to have a positive indication of the pending low inflows to the storage lakes and management would have been in a better position to take decisions which would have led to less severe effects of the drought.

The approach circumvents the disadvantages of inflow estimations associated with the physical characteristics of the catchment and facilitates maximisation of the available information relating to the catchment in terms of historic inflows, historic weather data and circulation indices. It is also flexible in terms of the ability to use it as a "one-stop-shop" for both short-term and long-term forecasting. The weakest link in the study is the forecasting of annual or seasonal precipitation amounts. However the study provides an operational framework for making reasonable forecasts of inflows to the storage lakes. Further research on particular elements of the integrative approach can improve the capability of the approach without any significant change to the framework.