Determining safe yield of an existing water supply is a basic aspect of water supply planning. Where water is withdrawn from a river directly without any storage, the withdrawal is constrained by the worst drought flow in the river. There is no flexibility for operational adjustments other than implementing conservation measures. Where there is a storage reservoir, yields higher than the flow in the source stream can be maintained for a period of time by releasing the water in storage. The determination of safe yield in this situation requires elaborate computation.

This thesis presents a synthesis of methods of drought flow analysis and yield estimation. The yield depends on both the magnitude of the deficit and its temporal distribution. A new Markov chain analysis for assessing frequencies of annual flows is proposed. The Markov chain results compare very well with the empirical data analysis. Another advantage of the Markov chain analysis is that both high and low flows are considered simultaneously; no separate analyses for the lower and upper tails of the distribution are necessary.

The temporal distribution of drought flows is considered with the aid of the generalized bootstrap method, time series analysis, and cluster sequencing of worsening droughts called Waitt's procedure. The methods are applied to drought inflows for three different water supply reservoirs in Spotsylvania County, Virginia, and different yield estimates are obtained.