Generating Forecasts of Reservoir Storage Using Climate Information from Probability of Exceedance (POE) Plots of Seasonal Rainfall

There has been much progress in improving the skill and reliability of seasonal rainfall and temperature forecasts for particular seasons and particular regions of the globe over the last twenty years. Advances are currently underway to develop capacity to generate skillful rainfall and temperature forecasts at the sub-seasonal timescale. How does a reservoir operator use a seasonal or sub-seasonal rainfall forecast, typically provided as probabilistic forecasts indicating whether an upcoming season will be above-, near- or below-normal, to decide on whether to implement a drought contingency trigger for a given reservoir? The answer is not always clear. A confounding factor is that a probabilistic forecast heavily weighted (e.g. 60‒70% probability of below-normal) say towards the below-normal category does not directly translate as the expected rainfall for the upcoming season being 60‒70% less than normal.

An alternative is to use the deterministic forecast of rainfall (i.e. the point forecast of the actual quantity of rainfall), in conjunction with an exceedance probability curve of historical observed rainfall, for the area of interest to obtain the exceedance frequencies for the point forecast and for the upper and lower limits of the point forecast using a confidence interval of choice.

We report on how we have developed a framework by which reservoir operators in Texas could consult seasonal rainfall forecasts, expressed as probability of exceedance plots, to develop short term (up to three months ahead) reservoir storage forecasts. These storage forecasts can be used to inform the implementation of drought contingency triggers for water supply reservoirs in the state.