The Value of Forecasts in Managing Extreme Events

Forecasts of future weather provide valuable information for reservoir operations, particularly during floods or droughts. A challenge confronting reservoir operators today is whether to incorporate new climate products into their operations to help manage such extremes, or to use historic data to guide them, perhaps Ensemble Streamflow Predictions (ESP). This research evaluates the accuracy and value of forecasts generated from the Climate Forecast System version 2 (CFSv2) using the operations of Bear Lake, a multi-purpose reservoir owned by Pacific Corps, and compares it to the accuracy and value of using an ESP approach. Streamflow reforecasts are generated and used to evaluate the predictive skill of the CFSv2 in reservoir management. Using the CFSv2 may offer more insight when responding to climate driven extremes than the ESP approach, because the CFSv2 approach incorporates the current state of climate into its forecasts rather than using all of the historic record as being equally probable.

For the Bear Lake system, located on the Utah/Idaho border, forecasts are most critical during the April through September period, when releases are being made for irrigation. Since Bear Lake provides for a multi-year draw-down, forecasts provide the most value when storage is unusually high or low. Snowpack data, available from April to June, are a determining factor in streamflow runoff during the later spring and early summer. The CFSv2 reforecast data makes use of this information and the approach used this research also uses snowpack data to select appropriate analog years from the ESP dataset. The streamflow forecasts are used as input for a decision support system.

The decision support system developed for this study includes a simulation model that incorporates system constraints and operating policies. To determine the value of the reforecast products, performance metrics meaningful to managers are identified and quantified. Without such metrics and awareness of seasonal operational nuances, it is difficult to identify forecast improvements in meaningful ways. Some of the important operational metrics formulated for the Bear Lake Project are maximizing release irrigation allocations and reliably providing set allocations. The analysis of these metrics focuses on high storage scenarios where forecasts are used to effectively manage to prevent floods, and low storages scenarios when managers consider cutting back on releases. These metrics of system performance are compared for the reforecast, climatology, and observed scenarios to evaluate the potential benefits of using CFSv2 seasonal forecasts in systems decision making.