Relative Skill of IFS Models

It is important to have measures of both forecast skill and relative differences in error of ensemble control, ensemble mean, and individual ensemble members.  This allows the forecaster to assess the strength of one product over another and the way this varies through the forecast period.  The Anomaly Correlation Coefficient (ACC) and the Equitable Threat Score (ETS) show the comparative performances on a global scale.   The relative differences in error with lead-time shows comparative errors with lead-time.  Relative weights can be assigned to the ensemble at different lead-times to better use the strengths of each in a more structured way.

Ensemble Control Forecast might be incorrectly thought of as being a deterministic model yielding deterministic results.   But the unperturbed Ensemble Control Forecast is only one member of the ensemble.  There are 50 other (perturbed) members of the ensemble.  All are equally valid.  There is no reason to select the results from the unperturbed control member rather than any of the others.  Also any member viewed in isolation cannot provide any estimate of forecast uncertainty or confidence.   


(FUG associated with Cy50r1)