A stochastic hydrometeorological model for the optimization of multi-reservoir operation

Abstract

A hydrometeorological model is presented that utilizes 30-day meteorological forecasts of temperature and precipitation issued every 15 days by the U.S. National Weather Service to provide knowledge of the future hydrometeorological conditions of a river basin. The model is entitled "Monthly Operational Hydrometeorological Simulator (MOHS)." Use of the 30-day meteorological forecast categories of light, moderate, or heavy precipitation and below normal, near normal, or above normal temperature provide physical constraints upon quantitative values which were synthesized by a Monte-Carlo simulation technique. Monthly precipitation data for meteorological stations semi-randomly located within the river basin are simulated from the square-root-normal distribution, while monthly mean data of ambient-air temperature are simulated using the Gaussian distribution assuming a "perfect" forecast. Contingency tables of forecast versus observed weather for the 24 forecast periods per year were obtained. It was demonstrated that many forecast periods provided information that verified better than chance. The contingency tables also provided conditional probabilities for each forecast category and period. These then were used to calculate an "expected value" forecast of temperature and precipitation for each station by assuming an "imperfect" forecast. Both methods used an objective analysis scheme that fitted the station data to a rectangular grid-coordinate system. The analyses reproduced the rainfall and temperature patterns well and facilitated computations of surface runoff, reservoir evaporation, and consumptive-use by crops. ...