由不同的觀測反演估計、分析與模擬實驗檢視台灣鄰近區域降雨的長期氣候分布特性

The Observed and Simulated Climatic Rainfall Distribution over Taiwan and Surrounding Region with Different Retrieval Techniques, Analyses, and Numerical Experiments

陳正達(Cheng-Ta Chen)

《大氣科學》 28卷2期 (2000/06) Pp. 115-142

摘要

利用新近彙整分析的各種觀測降雨資料集與數值模擬結果，本文以台灣及鄰近區域爲研究焦點，比較不同降雨資料集所呈現的區域長期氣候分布特性。在區域平均年降雨量方面，以紅外線反演比其他觀測估計方法來得小，特別是在冬半年。緯度分布變化方面，靠近赤道的輻合帶降雨量比其他方法高，而副熱带區域比較低，隨時間變化方面，則是傾向於呈現較大的季節變化。微波放射的雨量反演比微波散射法所估計的雨量略高，增加了測站雨量計分析的綜合估計，使得區域平均年降雨量比只用衛星資料的估計大，而主要的差異是源自於春、夏季以及台灣與大陸東南沿海地區，顯示衛星資料低估區域的鋒面降雨。在不同的降雨氣候資料集比較方面，區域平均年降雨量，以Legates的降雨氣候值最大，Xie and Arkin與Jaeger的降雨氣候次之，而全球降雨氣候計畫的衛星雨量計綜合估計最小。Legates的降雨氣候較高的原因，主要是冬季在台灣東側沿海有相當明顯的雨帶，以及夏季在巴士海峽與台灣南部的對流性降雨中心。Jaeger的降雨氣候在夏季比衛星雨量計綜合估計多，同時偏南的熱帶降雨也較大，間熱帶輻合區降雨變化振幅低估，並且錯置最大振幅區於台灣與巴士海峽。Xie and Arkin的綜合降雨氣候分析與全球降雨氣候計畫的衛星雨量計綜合估計較爲相似，Xie and Arkin的降雨資料略高，主要是由於整個熱帶與副熱帶海洋上有較強烈的輻合降雨帶。
在降雨模擬方面，歐洲中長期預報中心再分析資料中的降雨季節變化模擬得相當好，可以掌握在熱帶降雨帶隨季節的移動時間與變化幅度，以及台灣鄰近區域在夏季七月份降雨的局部轉小。美國國家環境預報中心的再分析資料中，最大的問題在於低估台灣鄰近區域五、六月的平均雨量，導因於春夏轉換之際西太平洋副熱帶高壓的位置掌握有誤，同時期降雨較多的中心反而在大陸華南地區。ECHAM4氣候模式在間熱帶輻合區的對流性降雨量與季節變化方面，多半是高估的，而台灣附近的副熱帶降雨則略有不足，特別是在夏季。在台灣以北的鋒面降雨系統則完全未能掌握正確的季節相位變化，有提早在大陸華南、長江流域一帶產生較大降雨的趨勢，而模式的副熱帶高壓系統，未能在五、六月之際退出南海北部與台灣附近、移轉至西太平洋，造成該時期的區域降雨低估。模式水平解析度增加並無法顯著而系統化地改善模擬缺失，運用不同的深對流積雲參數化閉合方式，對於間熱帶輻合區的降雨季節變化有明顯的影響。在台灣鄰近地區降雨年際變化方面，模式模擬降雨距平與尼紐3區海溫距平，在1979至1993年期間的相關係數分布，與觀測相當接近，儘管區域平均月降雨距平的模擬表現，由於相關係數低，與觀測降雨距平不盡相同。

關鍵字

降雨分佈 ； 水文循環 ； 區域氣候

並列摘要

Using the various recent precipitation estimates, this study compares the long-term climatic precipitation distribution over Taiwan and surrounding area in the different observed and simulated datasets. For annual mean rainfall, the IR-retrieval is less than using other techniques, especially from October to April. IR-retrieval also estimates more rainfall in the Intertropical Convergence Zone(ITCZ)and less in the subtropics. The regional precipitation retrieval using microwave emission technique tends to be slightly larger than microwave scattering technique through the year. The impact of rain gauge measurements is to increase the precipitation estimated from satellite data. The main difference is in Taiwan and the southeastern China where satellite retrieval underestimates the precipitation rate during boreal spring and summer. For the different precipitation climate analyses, the Legates-Willmot rainfall climate analysis produces the largest annual mean regional rainfall. The area precipitation in the rainfall analyses by Jaeger and Xie and Arkin is smaller. The satellite-gauge combined precipitation estimate from Global Precipitation Climatology Project (GPCP) yields the smallest regional rainfall. The main reasons for the larger precipitation in Legates climatology are the rainband in the east of Taiwan during winter and the convective rainfall center over the southern Taiwan and Bashi Channel during summer. The Jaeger climatology produces more rainfall than the GPCP data in summer and over the tropical precipitation bands. Its amplitude of seasonal variation is underestimated and the location of the maximum center is mislocated northward. The rainfall analyses by Xie and Arkin is similar to the GPCP satellite-gauge combined estimate. The somewhat larger rainfall in Xie and Arkin's dataset is due to the more vigorous convective rainfall centers over the tropical and subtropical ocean.
For the rainfall simulation, the precipitation rate in the ECMWF reanalyses data has reasonable seasonal and latitudinal variations. It can also capture the detailed change of area precipitation in July. The problem of rainfall simulation in the NCEP reanalyses model is the underestimation of the precipitation near Taiwan in May and June.It is caused by error in the position of Pacific subtropical high. The major rainfallcenter is shifted to the southern China in this period.ECHAM4 climate model runs tend to overestimate the convective precipitation in the ITCZ. The subtropical rainfall is relatively small, specially during the summer. The simulated phase change of the seasonal variation of the frontal system rainfall north of Taiwan is incorrect. The model tends to predict more precipitation over the Yangtze river valley in the spring. The Pacific subtropical high in the model did not retreat from the South China Sea to the tropical Western Pacific in May and early June. This leads to the underestimation of regional rainfall during the period. The increase of model resolution in general can not significantly improve the model deficiency. The impact of different closures for the deep convection on the rainfall seasonal variation in ITCZ is apparent. For the interannual variability of regional precipitation, the distribution of correlation coefficient for the simulated rainfall anomalies and NINO3 SST anomalies from 1979 to 1993 is very close to the observation. Due to the relatively low correlation value, the area mean monthly rainfall anomalies is not captured well by the ECHAM4 model.

並列關鍵字

Rainfall distribution ； Hydrological cycle ； Regional climate