Forecast Technique

Mean benefits [$M/dec]

Mean losses [$M/dec ( f)]

Dependability [%]

Superior decadal [%]

Benefit cases annual [%]


Perfect Forecast

3,350



100%
  
prob = 80%
     
Statistical

2,740

25 (10%)

100%

18

23

Dynamical

2,610

100 (66%)

100%

0

15

Multimodel

2,780

5 (2%)

100%

82

35

NoForecast

2,610



> 99%

0

27

prob = 95%
     
Statistical

2,500

35 (23%)

100%

8

20

Dynamical

2,500

237 (23%)

100%

0

0

Multimodel

2,500

48 (5%)

100%

88

44

NoForecast

2,440



100%

4

36

 Table shows economic benefit over 10year time periods of reservoir management for a planned reservoir in the Blue Nile Basin of Ethiopia, based on assumptions of the upcoming seasonal climate predicted by a statistical model, a dynamical model, a combination of statistical and dynamical models, and NoForecast, which assumes that every year behaves the same. The probability is the confidence level from the prediction in formation in a given year that the precipitation, and thus inflow, will exceed a certain value. That value then triggers the release decisions and determines how much energy can be generated.
 Notation: “Prob” = precipitation probability of exceedance; “$ M/ dec” = million US dollars per decadal simulation; “f “ frequency of occurrence in%; “Dependability” = percent of months above threshold. “Mean losses” represents average of simulations for which loss occurred, defined as years when noforecast benefits are greater than given forecast technique benefits. For “superior benefit” cases, quantities reflect percent of simulations for which that technique produced benefits greater than other techniques. “Perfect forecast” uses observed precipitation; “noforecast” uses climatological precipitation. See (Block & Goddard 2012) for details on hydrological models and climate forecasts used to generate the table.