IntermtnClimateSummaryJan2006 - Laserfiche WebLink

Home Browse Search

FEATURE ARTICLE FROM INTERMOUNTAIN WEST CLIMATE SUMMARY, JANUARY 2006 (Continued from p.3) Wet G Gild I FGre.castPelforma flee Bad I . :!!!!!! NO DATA (Situation has not occured) Dry G Go-d I Bad I . :!!!!!! NO DATA (Situation has not occurecf) Ftxe.castPelforma flee Figure Ib: An example result of the Forecast Evaluation Tool. The False Alarm Rate results for climate forecasts issued in November for the December-February season. Southeastern Utah's and south-central Colorado's winter forecasts tended to be the most successful of all the climate divisions in Colorado Wyoming and Utah, especially for predicting drier than average conditions (map at right). For example, the 0 scores in those divisions indicate that every forecast for dry conditions on the last decade panned out. Forecasts for wet winters in the Intermountain West region only came to pass about half the time or less (map at left). and Ranked Probability Score) are more useful to understand the forecast perfor- mance. Probability of Detection This analysis indicates how often a forecast was made for non-average condi- tions compared to the total number of times it actually occurred. Your results will include separate maps for forecasts of above-average events (wet or warm) versus below-average events (dry or cool). To convert the resulting climate division score into a percentage, just multiply the resulting value by 100. A score of 0.346 for detecting wet conditions for the selected season means the CPC issued a forecast calling for above-average precipi- tation in about 34.6 percent of the cases in which precipitation tallies registered as above-average. Emergency managers have indicated they find these scores useful. Ranked Probability and Brier Scores While the Brier score differentiates categories into wet and dry (or warm and cool), the Ranked Probability score provides one lumped result for both condi- tions. Other than that, they have similar features. Both scores take into consider- ation the strength of the issued forecast. So, if above-average conditions prevail as the CPC had predicted, a forecast issued with a 70 percent probability gets a higher score than one issued with a 40 percent probability. Similarly, the 70 percent prob- ability forecast takes a bigger penalty than the 40 percent probability if conditions turn out to be average-and an even big- ger hit if conditions turn out to be below- average. The Brier and Ranked Probability skill scores represent the proportion of time above and beyond what would be expected by chance (3 3 percent). That's partly why a climate division with a Prob- ability of Detection score of 0.517 can translate into a Brier skill score of 0.086. This also explains why some of the skill scores turn up negative, indicating the viewer theoretically could have done bet- ter just by flipping a three-sided coin. Customize your options Now you have the know-how to consider how forecasts fare during a variety of sea- sons with a number of different lead times, using evaluation approaches that suit your needs. The website has many other features to explore on your own. Want to know more? If you have any questions about how the website works, you can send an email to: hydis_team @hwr.arizona.edu. Support for development and imple- mentation of the Forecast Evaluation Tool came from the National Oceanic and Atmospheric Administration, the NOAA-junded Climate Assessmentfor the Southwest (CLIMAS) and GEWEX Americas Prediction Project (GAPP) programs, the National Aeronautical and Space Administration, NASA's Hydrologic Data and Information System (HyDIS), EOSDIS Synergy programs, the National Science Foundation, and the NSF-junded Semi-Arid Hydrology and Riparian Area (SAHRA) Science and Technology Center. FEATURE ARTICLE I 4/~