Making optimal use of climate forecast uncertainty in water resources operations and management

The water management sector is often thought of as a primary beneficiary of improved climate forecasts. Given the probabilistic nature of climate forecasts, however, the potential impact of uncertain climate information on, say, reservoir decisions are not well understood. A simple statistical framework is used here to understand the conditional forecast distribution of runoff volume based on a climate forecast. Three contributing factors are examined: a. the correlation between the climate precursor (say, El Nino-Southern Oscillation) and the target variable (here, runoff volume), b. the forecast uncertainty itself, which can be strongly case-dependent, and c. the uncertainty and temporal variations in the teleconnection metrics. The impact of forecast-type (climatology, climate information with no uncertainty, and climate information with uncertainty) on the runoff volume distribution is quantified for a simple reservoir system. Knowledge of the forecast uncertainty in runoff volumes, together with reservoir operation constraints, enables an a priori identification of cases when the probabilistic forecast is likely to be most or least useful for a particular reservoir system.