FIRM RESERVOIR YIELD—HOW RELIABLE ARE HISTORIC HYDROLOGICAL RECORDS?

Abstract

Reservoirs are of necessity always built on the basis of incomplete hydrological information which introduces uncertainty into their design and operation. Since the advent of the electronic digital computer attempts have been made to reduce the uncertainty in hydrological design and reservoir management by the use of synthetic hydrology and simulation. It has been found by simulation that the expected benefits from a proposed reservoir system are often a function of the stochastic process selected for the synthetic hydrology, as well as depending upon the magnitude, and choice of driving parameters (commonly, the mean, variance, lag one serial correlation and Hurst's ‘h’). It is suggested that hydrological records are often two short and most statistical tests too weak for the hydrologist to be able to pick ‘the correct’ synthetic hydrological world with any reasonable degree of certainty. However, it would appear that for many problems and places that there is sufficient hydrological data for the hydrologist to assign probabilities to various prior distributions, and to optimize reservoir management and design by Bayesian decision theory.