系集機率擬合平均定量降水預報產品之分析：以2014年梅雨季為例

The Characteristics of the Probability Matched Mean QPF for 2014 Meiyu Season

蘇奕叡(Yi-Jui Su) ； 洪景山(Jing-Shan Hong) ； 李志昕(Chih-Hsin Li)

《大氣科學》 44卷2期 (2016/06) Pp. 113-134

摘要

系集預報的目的是要包含預報的不確定性，以補單一模式預報的不足；然而，如何從龐大的系集預報資料中產製有用的預報產品，此為重要的課題。本研究評估以中央氣象局區域系集預報系統為基礎所發展之機率擬合平均(Probability MatchedMean, PMM)降水產品的特性與預報表現。研究結果顯示PMM定量降水預報產品，可有效改善系集平均定量降水預報對大雨低估的情形，並且有比決定性預報更高的可預報度。然而，校驗結果亦顯示PMM降水產品對大雨有過度預報的情形，此一現象無法經由機率擬合方法之修正而獲得足夠的改善，顯示對於大雨過度預報情形為模式本身的系統性偏差所致。此外，為因應不同氣象資料使用者的需求，可發展出不同累積區間的PMM產品，但分析顯示使用較短累積區間的PMM加總成較長時間的定量降水預報會強化大雨的極值，此一現象隨著累積區間之縮短而更為明顯。

關鍵字

機率擬合平均 ； 系集預報

並列摘要

Ensemble forecast is expected to cover the uncertainties due to the deficiency of the limited predictability from the single forecast. However, how to derived useful forecast product from the large ensemble dataset is of most important and challenging. This research aimed at the assessment of characteristics and performance of the Probability Matched Mean (PMM) rainfall product derived from the operational ensemble prediction system in Central Weather Bureau. The results show that the PMM is able to re-construct the extreme rainfall as the observations, to improve the under-estimate of the ensemble mean, and provide better predictability against the deterministic forecast. However, the PMM also shows the slightly over-prediction at the larger rainfall threshold. The over-prediction was due to the systematic bias from the ensemble prediction system and can’t be removed from the fine-tune of the PMM algorithm. For the practical application, PMM can be designed according to the different accumulation period. This study pointed out that the accumulated rainfall from the PMM with shorter accumulation period will result in the enhancement of the extreme rainfall.

並列關鍵字

Probability Matched Mean ； Ensemble Forecast