Design drought should be estimated in a probabilistic framework. A method is presented for estimating the probability and the return period (recurrence interval) associated with the design drought of a water resources project such as a treatment plant, irrigation scheme or a low flow augmentation system. Assurance is the probability of nonoccurrence. Sum of assurance and permissible risk equals unity. Return period of the design drought is a function of the project’s useful life and the desired level of assurance.

Probability distribution that describes the data of droughts must be established for the computation of design drought and the risk associated with it. An analysis of the annual minimum streamflows in Missouri River Basin was performed to establish a suitable probability distribution. The annual drought series of 37 stations were selected for analysis on the basis of adequate length of record, freedom from upstream diversions and regulations, and randomness of data. Tests of goodness-of-fit were performed on the gamma, the log-normal, and the Weibull distributions for each series. Comparison of the results of the tests revealed that the gamma distribution was the best of the three distributions tested.