Performance transparency and responsible brokerage of seasonal forecasts
Robert E. Livezey, OCWWS/NWS/NOAA, Silver Spring, MD; and M. Timofeyeva, B. Mayes, M. J. Brewer, and F. Horsfall
Federal and federally supported institutions that advocate or promote the use of climate forecast products, either directly or through research into and development of applications of these products, have a basic responsibility. Their brokerage of these products needs to be honest and neutral. To do this they must understand (to the extent possible) the performance characteristics and limitations of these products and make every effort to convey this knowledge transparently to potential users, including matching this knowledge with user attributes in developing and promoting applications. Relevant attributes are user interests, sophistication, frequency and consistency of use, and tolerance for error. Very often these mandatory considerations are ignored, most often a consequence of well-meaning naiveté. Nevertheless, both product credibility and users are placed unnecessarily at risk. Because of their growing visibility and popularity, this situation is especially serious for seasonal forecasts.
Highly successful seasonal forecasts are made only on an opportunistic basis (temporally, spatially, or both) and far from routinely. Otherwise these forecasts are expected to have either no or only very modest skill. There is only one opportunity each year for any particular season/lead/location to base a decision on a forecast. These circumstances are frequently ignored in associated outreach and application activities and in value studies. Coupled with the lack of sophistication of many user groups, the result is misleading at best. More seriously it may lead to user losses from inappropriate reliance on a forecast or forecasts. We will illustrate many of these points by a combination of arguments, including robust seasonal prediction verification statistics, the implications of infrequent and/or undisciplined use on the implications of cost/loss vs. value models, and the practical interpretation and limitations of commonly used skill measures..
Session 3, Applications of Seasonal Predictions
Tuesday, 21 June 2005, 1:30 PM-2:15 PM, South Ballroom
Previous paper Next paper
Browse or search entire meeting
AMS Home Page