台灣氣候變遷推估研究

Projection of Climate Change Over Taiwan Using Statistical Downscaling Scheme

陳正達(Cheng-Ta Chen) ； 朱容練(Jung-Lien Chu) ； 許晃雄(Huang-Hsiung Hsu) ； 盧孟明(Mong-Ming Lu) ； 隋中興(Chung-Hsiung Sui) ； 周佳(Chia Chou) ； 翁叔平(Shu-Ping Weng) ； 陳昭銘(Jau-Ming Chen) ； 林傳堯(Chuan-Yao Lin) ； 鄭兆尊(Chao-Tzuen Cheng) ； 吳宜昭(Yi-Chao Wu) ； 卓盈旻(Yin-Min Cho) ； 陳重功(Jonq-Gong Chen) ； 張雅茹(Ya-Ju Chang) ； 林士堯(Shi-Yao Lin) ； 林修立(Shiou-Li Lin) ； 童裕翔(Yu-Shiang Tung) ； 楊承道(Cheng-Dao Yang)

《大氣科學》 42卷3期 (2014/12) Pp. 207-251

摘要

本研究以全球氣候模式模擬的結果為基準，運用統計降尺度技術，分析未來臺灣氣候變遷的可能性機率分布範圍。運用機率方式表示是以科學與客觀的方式呈現上述氣候模式差異、自然氣候變動以及降尺度方法所疊加的不確定範圍，同時區域的氣候變遷推估機率分布特徵也可以用以評估我們對未來臺灣氣候變遷推估的信心度。以A1B單一未來發展情境為例，下列是針對21世紀末未來臺灣氣候變遷推估的主要發現：（1）在近地表氣溫季節及區域平均的變化部份，所有氣候模式所推估中位數大約是介於2.1°C至2.4°C之間，北臺灣較南臺灣的增溫幅度略高，而秋季較其他季節略低。約有一半的模式顯示未來變化主要是介於1.8°C至2.9°C之間。90%以上的模式推估區域季節平均變化將在1.3°C以上或小於3.1°C，同樣是北臺灣較南臺灣的增溫幅度略高。（2）在季節及區域平均降水的變化方面，未來冬季雨量有近乎四分之三的模式推估變化都是減少的，約有一半的模式的推估是介於+0%至–23%之間，只有少數（10%）的模式推估區域冬季雨量減少可能在–34%以上或反而可能有+6%的增加。而未來夏季雨量變化方面，則有將近60%的模式推估降水都是增加，約有一半的模式認為未來夏季變化主要是介於+15%至+45%之間，雖然只有少數（10%）的模式推估區域夏季雨量增加可能在+34%以上。這些推估結果對原本就是夏季多雨、冬季少雨的中南部地區而言，如何調配豐枯水期的水資源，在未來將會是更為嚴峻。對於未來臺灣近地面氣溫變遷推估而言，無論是月、季或是年平均的氣候平均氣溫變化，在21世紀末的變化幅度都遠超過模式所估算的氣候自然變動與統計降尺度方法的誤差範圍。不過，對於未來臺灣降雨變遷推估的應用，必須注意其變化幅度往往未必具有統計上的顯著性，模式推估分布的可能性機率分布範圍或甚至符號都改變，科學上對相關結果可靠性的信心則相對較為不足。

關鍵字

統計降尺度 ； 不確定性 ； 氣候變遷 ； 氣候變遷推估

並列摘要

In this study, statistical downscaling method is applied to the climate projections simulated by 24 General Circulation Models (GCMs) to investigate the climate change in Taiwan in the end of 21st century. The probabilities of changes in temperature and precipitation are calculated and used to objectively present the uncertainty caused by the differences between models, natural variability, and downscaling methods. Meanwhile, the characteristics of natural variability are also used to evaluate reliability of climate change projection in Taiwan. Based on the climate projections of A1B scenario of 24 GCMs, the primary findings are: 1. For the seasonal and regional mean surface air temperature, the median of temperature increases are between 2.1°C to 2.4°C, larger over northern Taiwan, smaller over southern Taiwan, and smaller in autumn. Half of projections indicate the temperature increases are between 1.8°C to 2.9°C; 90% of projections show the temperature increases are larger than 1.3°C; and 90% of projections show the temperature increases are smaller than 3.1°C. 2. For the seasonal and regional precipitation, winter precipitation in 75% of projections decreases and the decreases range from +0% to -23% in half of projections. Only less than 10% of model projected future winter precipitation reduction exceeded -34% or even increase rainfall by +6%. The summer precipitation in nearly 60% of projections increases. There are about half of model projected summer precipitation increases with range from +15% to +45%. Only less than 10% of model projected future summer rainfall change exceeded +34%. These results implied that the issue of water resources allocation for central and southern Taiwan, where winter is the dry period and summer is the wet period, would become more challenging in the future. Moreover, the warming signal in surface air temperature, which is much larger than the natural variability of temperature and the error caused by statistical downscaling methods, is significant and should be seriously considered. On the other hand, it should be noted that the changes in precipitation may not be statistically significant in many occasions.

並列關鍵字

Statistical Downscaling ； Uncertainty ； Climate Change ； Projection