The Cloud-resolving Storm Simulator (CReSS) has been used by the author to produce forecasts at real time for Taiwan, at a high-resolution of 2.5 km since 2010. As a contributing member in the Taiwan Cooperative Precipitation Ensemble Forecast Experiment (TAPEX) of the Taiwan Typhoon and Flood Research Institute (TTFRI), these forecasts are carried out four times a day (at 0000, 0600, 1200, and 1800 UTC) for 72 h (or 78 h) and posted on the author s website at http://vortex.es.ntnu.edu.tw/ once available. The domain was 1080 km by 900 km in 2010-2011 and has been enlarged to 1500 km by 1200 km since 2012. Being the only cloud-resolving member in TAPEX, the 2.5-km CReSS has the highest resolution and largest fine-domain size among all models in real-time operation in Taiwan.
Through CReSS forecasts for typhoons in Taiwan in 2010-2012, a strong dependency of model performance in quantitative precipitation forecasts (QPFs) as measured by skill scores such as the threat score (TS) on rainfall amount, i.e., the better model performs when there is more rain, is demonstrated. Implied simply from the positive correlation between rain-area sizes and scores, the dependency is also attributed to the model s capability to properly handle (within 72 h) the processes leading to more rain, which are largely controlled by the typhoon s track, size, structure, and environment and Taiwan s topography. The phenomenon is thus expected to exist in other models, rainfall regimes, and regions around the World.
Due to this dependency, the skill of model QPFs for extreme events can be assessed accurately only when forecasts targeted for periods of comparable rainfall magnitude are included for averaging. For the most-rainy 24-h of the top-5 typhoons, the 0-24-h QPFs by CReSS have mean TS of 0.67, 0.67, 0.58, 0.51, and 0.32 at thresholds of 25, 50, 130, 200, and 350 mm, and 0.64, 0.57, 0.37, 0.33, and 0.22 from 48-72-h QPFs, respectively, suggesting superior performance even 2-2.5 days in advance. These scores are strikingly high, precisely because Taiwan can receive extreme rainfall from typhoons. For smaller (non-hazardous) events, the mean scores are progressively lower, but also unimportant and less representative statistically. Therefore, it is inappropriate to average scores over multiple forecasts as those from less-rainy periods would contaminate the result for key periods. The implication for forecasters is also discussed.