Poster Session P1.25 The development of forecast confidence measures using NCEP ensembles and their real—time implementation within NWS web—based graphical forecasts

Monday, 1 August 2005
Regency Ballroom (Omni Shoreham Hotel Washington D.C.)
Andrew V. Durante, Florida State University, Tallahassee, FL; and R. Hart, A. I. Watson, R. H. Grumm, and W. Drag

Handout (115.3 kB)

Ensemble model data can provide a wealth of knowledge to forecasters especially in terms of forecast confidence. A model run where members diverge corresponds to a low confidence forecast while a model run where members converge corresponds to a forecast of high confidence. The current NWS graphically based forecasts accessible to the public do not show this measure of uncertainty and thus communicate an often exaggerated sense of precision and confidence.

Starting last August and extending into 2006, analysis of about 3 years of individual global GFS ensemble data is occurring as part of a COMET cooperative project with the NWS Office in Tallahassee. A climatology for each GFS ensemble member is being developed as a function of variable, location, time of year, and forecast length. Once the normalized climatology distributions are calculated, forecast confidence/uncertainty measures can be developed from comparing the normalized spread of the real-time GFS ensemble members to the average spread of the GFS ensemble climatology. This normalized spread will also be compared to the typical spread for that time of year and location to arrive at a relative measure of forecast uncertainty. If the current model ensemble uncertainty is greater (less) than the uncertainty of the model ensemble climatology, then there is a lower (higher) than normal confidence.

Using the first year of GFS ensemble data, preliminary confidence graphics have been developed and analyzed ( to examine how confidence values behave within certain synoptic situations. Preliminary results have shown that when the forecast confidence is above normal, the average NWS forecast error is approximately 40 to 50% of the error when the forecast confidence is below normal. For example, in the New England states on 7-8 March 2005, a strong coastal storm had areas of below normal confidence associated with the storm track and areas of above normal confidence behind the storm track. Forecasts of high and low temperatures made 3 days prior to the storm busted by as much as 22 degrees while the 5 day forecast issued on the same day had errors of 3 to 4 degrees on average. Several case studies will be examined in this paper to illustrate how the forecast confidence graphics can be used to not only speed up the forecast process for the forecaster, but also to improve the accuracy of resulting forecasts by knowing when using climatology for a forecast is most appropriate.

Supplementary URL:

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner