Stochastic Hydrologic Model for Drought Management

Abstract

A method of producing long-range water-supply forecasts is developed and applied. The method is capable of producing conditional cumulative distribution functions (CDF) for monthly streamflow volumes and producing a set of equiprobable future streamflow traces. Using the latter output, it is possible to evaluate the conditional CDF of reservoir storage for any particular future month, or probabilities of storage falling below some level during the forecast period. The model used in an ARMA (1,1) cyclic model. It differs from an ARMA (1,1) stationary model in that the correlation structure varies with time of year. The model is fit in this analysis to the 81-year long stream-flow record of the Potomac River at Point of Rocks, Maryland.