臺灣極端氣溫之季節潛在可預報度評估和氣候暖化的影響

Evaluation of Seasonal Potential Predictability of Temperature Extremes in Taiwan and the Influence of Climatic Warming

The daily weather variability is considered as a noise for the seasonal climate prediction and thus its influence needs to be alleviated as much as possible. This study proposes a numerical method that relaxes the assumptions normally made in weather noise estimations within a season, thereby likely increasing seasonal potential predictability (SPP). The proposed numerical method is first applied to the centennial de-trended monthly-mean surface temperature maximum (Tmax) and minimum (Tmin) data in Taiwan's measurement stations. The numerical solutions show that both the monthly mean variances of weather noise, not necessarily modeled by linear/stationary assumptions, and their inter-monthly correlations loosen the stationary assumption or register as non-zero values (if one month apart or even more, e.g. between January and March). Compared to the analytic solution of simplified equations, the numerical solutions in this studyg enerally have much better coherencies with those estimated directly using the daily data. The estimated SPP of Tmaxis slightly lower in winter but higher in other seasons. The estimated SPP of Tminis slightly lower in fall and winter but higher in spring and summer. The overall means in both Tmax and Tmin(47.6%) are comparable to the estimations of the analytic solution (46.2%), but significant differences appear at individual stations and seasons. The warming trend effects on the SPP are examined by applying the new method to the trend series. The results showed that coherency between a warming trend and an increased SPP is significantly higher in Tmin than in Tmax. On average, the SPP of Tmin in summer (winter) is 86.3% (66.8%), and 22.7% (20.3%) higher than that of the de-trended SPP estimations. The overall mean of the trending SPP of Tmin is around 75%, a 27.4% increase compared to the de-trended SPP. Conversely, the overall mean of the trending SPP of Tmax (48.5%) is similar to that of the de-trended estimate. The all-season regression shows that the warming efficiency, which is defined as the linear slope of the changed SPP (trend-remained minus de-trended) regressed on the trend values of seasonal-mean station temperature extremes, shown an increase of around 16% (2%) per 1.0 ℃ /100-yr warming for Tmin (Tmax). Conclusively, the climatic warming is an external source of SPP for temperature extremes, especially for the Tmin. 對季節氣候預測而言，每天的天氣變化是一種干擾，必須儘可能濾掉其影響。本研究提議用一種數值方法可鬆弛掉對季節內天氣噪動常用的假設，進而或可提高季節的潛在可預報度（Seasonal Potential Predictability; SPP）。提議的方法首先被應用於台灣百年長度，去掉線性趨勢的測站月平均地表最高溫（Tmax）和最低溫（Tmin）。不必然需要假設為線性或定常才能模擬的天氣噪動的月平均變異量，以及其在月與月之間的相關係數，一旦放掉通用的定常假設或註記為非零值（假若相隔在一個月或以上的話，比如介於1月和3月），則數值解和直接使用日資料所得到的估計值之間的一致性，與簡化方程下的解析解相比，將會好得很多。Tmax的SPP估計值，除了冬季略低之外，其他季節都較高。Tmin的SPP估計值則是在秋冬兩季略低﹔但春夏較高。Tmax和Tmin的總體平均的SPP估計值（47.6%）約與解析解相當（46.2%），但兩者在個別測站和不同季節呈現有明顯差異。新方法被用到線性趨勢保留的原氣溫序列以便檢查氣候暖化對SPP的影響。結果顯示暖化的上升趨勢和增加的SPP間的一致性，Tmin明顯高於Tmax。平均而言，夏季（冬季）Tmin的SPP可達86.3%（66.8%），比原先去掉線性趨勢得到的SPP高出22.7%（20.3%）。其四季總體平均的SPP估計達75.0%，與去掉線性趨勢後相比，增加了27.4%。相反地，保留線性趨勢的Tmax其SPP的總體平均值（48.5%）則與先去掉線性趨勢後估計的SPP值，約略相等。不分季節的迴歸結果顯示，暖化效率，定義為改變的SPP（保留線性趨勢減去去掉線性趨勢）與季節平均測站極端氣溫序列的線性斜率，Tmin（Tmax）每百年增加1.0℃時，SPP大約增加16%（2%）。確實地，氣候暖化對於極端氣溫的SPP，尤其是Tmin，是外在的來源。