A comparison study between the use of single-point versus areal-mean rainfall values from a high-resolution model when verified against satellite retrievals for three typhoons hitting the Philippines

In this study, effects of using areal-mean instead of single-point output values from a cloud-resolving model to verify its quantitative precipitation forecasts (QPFs) in terms of the categorical 2 × 2 matrix against satellite rainfall retrievals are assessed for three typhoons that hit the Philippines. The motivation behind this work stemmed from the fact that categorical measures are point-to-point methods but satellite estimates are considered areal-mean values. Overall, we find that using areal-mean values have small but positive impacts on categorical statistics, mainly due to a better overall agreement between satellite-derived rainfall and model outputs after areal averaging with a smoothing effect. These impacts are also statistically significant up to high thresholds of roughly ≥ 350 mm. Using areal means from the model, the threat scores (TSs) improve at low thresholds by about 0.02–0.05, primarily owing to increases in the probability of detection (POD) of observed rainfall events, so the smoothing effect helps convert some misses into hits. In cases with low frequency bias, similar improvements in TSs also occur across middle and even high thresholds (up to 500–750 mm) as both POD and success ratio (SR) rise, with the latter indicative of a lowered false alarm ratio (FAR). Toward the extreme thresholds, results are more mixed and the confidence level of significance drops, but the TSs there are already low (≤ 0.08) against the satellite data regardless of the method adopted.