臺灣冬季溫度季節預報的預報因子挑選及可預報度來源分析

本文提出簡單但有物理根據的臺灣冬季（12-2月）溫度統計預報模式。以全球海表面溫度（SST）、陸地上2公尺氣溫（T2M）、海平面氣壓（SLP）三個氣候變數為範疇，經過繁複的預報因子篩選過程，確認出三個有預報技術的統計預報模式，繼而以2002－2017年冬季回溯預報實驗的預報結果說明這些模式的三分類預報命中率可作為選擇適合預報臺灣冬季溫度的季度氣候預報模式的判斷基準。三個統計預報模式的五個預報因子分散在四個不同區域：熱帶東太平洋、熱帶中印度洋、歐亞邊界的烏拉爾山、日本南端至南海北部的臺灣附近西太平洋海域，預報因子的變數有秋季平均SST、SLP以及秋季轉入冬季時SST、T2M變化幅度。預報因子與大尺度氣候變異遙相關模態的對應關係顯示臺灣冬季溫度年際變化可預報度根源於太平洋與北大西洋秋冬季的海表面溫度變異，但設想的情境能否實現則決定於大氣的遙相關反應，其中聖嬰現象（ENSO）、東大西洋與西俄羅斯模態（EA/WR）、西太平洋模態（WP）是聯繫海溫變異與臺灣附近冬季溫度變化的重要模態。預報單位若加強監測分析這些模態和海溫變異在入冬之前的關係，將有助於判斷臺灣冬季溫度的變化趨勢。

關鍵字

東亞冬季季風；季節預報；臺灣氣候

並列摘要

This paper presents a physically motivated empirical model (P-E model) for predicting Taiwan winter seasonal temperature with 0.5 month prediction lead time. The predictors are selected from three variables of global grid climate data, namely, sea surface temperatures, 2-meter air temperature, and sea level pressure. The predictor selection procedure is designed for reducing the influence of decadal-scale climate variability on the prediction performance. Through a retrospective forecast experiment, three P-E models with the best performance are selected to compare with the forecast skill of CWB 2-tiered monthly and seasonal dynamical forecast system TCWB2T2. The hit rate and Gerrity Skill Score are used as a baseline for evaluating the forecast performance. The procedure can actually be applied to any other seasonal forecast model for improving Taiwan winter temperature forecast. Through examining the relationship between the predictors and the influential large-scale teleconnection patterns to East Asian winter monsoon, we found the predictability sources of Taiwan winter temperature may have their root in SST anomaly over North Atlantic and Pacific through the teleconnection patterns, in particular ENSO, EA/WR and WP triggered by the anomalous SST. More studies are needed to evaluate the speculation.

並列關鍵字

East Asian winter monsoon ； seasonal climate forecast ； Taiwan climate

國際替代計量

臺灣冬季溫度季節預報的預報因子挑選及可預報度來源分析

全文下載

主題瀏覽