QPF Verification Information

This verification table shows statistics for 12-hr Quantitative Precipitation Forecasts (QPF) valid at 1200 and 0000 UTC from various numerical weather prediction models and manually derived value added forecasts from three different National Weather Service forecast centers; California Nevada River Forecast Center (CNRFC), the Weather Prediction Center of NCEP (WPC), and the Eureka and San Francisco Bay Area Forecast Offices. These models and forecasts are defined in the table below. The statistics provided are derived from a contingency table showing 12-hr forecast (QPF) and observed 12-hr precipitation (QPE) divided into 7 categories. If the forecast and observed 12-hr rainfall fall in the same category this is a hit. Otherwise this is a miss. The % correct, % to wet and % to dry are provided. The verification table shows the best guidance for the given duration of forecast selected, i.e. .5 days to 60 days, plus the next two best guidance and finally the worst guidance. Depending on the score chosen, color coding is used to show the quality of the particular model compared to the other forecasts. (green- top third, white- middle third, red- bottom third). If a duration longer than one-half day is chosen, a list of the wettest days over the watershed and the selected models performance is provided up to the ten wettest days. Missing forecasts are noted with dashes. Since the CNRFC tags all forecast with a 12z issuance time there will be no verification when the valid time is for 0000Z.

Threat Score (TS): A verification measure of categorical forecast performance equal to the total number of correct event forecasts (hits) divided by the total number of event forecasts plus the number of misses (hits + false alarms + misses). The TS is not affected by the number of no-event forecasts that are not observed (correct rejections). For this verification .5”/12hrs or greater precipitation is the threshold being verified.

Heidke Skill Score (HSS)¹: A skill corrected verification measure of categorical forecast performance similar to the Percentage correct (PC) but which takes into account the number of hits due to chance. Hits due to chance are given as the event relative frequency multiplied by the number of event forecasts.