Use of Stochastic Hydrology in Reservoir Operation

Intelligent operation of a complex reservoir system requires that the most current mathematical techniques be incorporated into the decision‐making process. One such technique is stochastic hydrology and, in particular, the use of a number of stochastically generated sequences (or traces) to provide a broader basis for management decisions during drought periods. This paper presents a case study illustrating the use of this approach by the Bureau of Reclamation in management of the Central Valley Project in California. A series of twenty 1,000‐yr stochastic traces for the Sacramento River Basin was generated using Reclamation's LAST (Lane's applied stochastic techniques) computer package, The package is described in Lane and Frevert (1988). Each trace was compared to an historically based natural flow sequence and was found to preserve adequately key statistical properties. The traces were then analyzed by use of a computer program which counted the number of occurrences of a flow below a specified level over 1‐ through 10‐yr intervals. Results for the twenty 1,000‐yr traces provided a much more objective estimate of nonexceedance probabilities for these low flows than might have been obtained by simply plotting the computed natural flows corresponding to the historic record itself.