本文分析中央氣象局全球預報模式（GFS）T42L18版本所執行之10樣本系集氣候（1950-2000）實驗，探討GFS對臺灣夏季氣候變化之模擬能力，重點包括模擬准確性與可預報度，並討論造成GFS氣候模擬中系統性誤差與影響可預報度高低的可能成因，及海溫變化於上述模擬特性中所扮演之角色。在準確性方面，GFS能合理模擬臺灣地區溫度之年際變化，對長期暖化現象亦能合理掌握，係由放GFS順應臺灣鄰近海域之SST變化來模擬臺灣地區溫度變化。GFS之主要系統性誤差為對臺灣地區降雨模擬出大致相反相位的錯誤變化，其原因為GFS因應海溫變化所主導的熱力與動力過程來模擬大氣降雨，但實際顴測中，臺灣鄰近海域係由大氣降雨變化來影響海溫變化，由於GFS實驗對此區域內之海洋－大氧交互作用模擬出相反相位，故模式中臺灣地區降雨亦呈相反相位變化。在可預報度方面，當模式內臺灣鄰近地區為偏暖變化時，臺灣鄰近海域海溫距平強度為影響臺灣夏季溫度可預報度高低的主要因素，強度較強（較弱）海溫距平，導引較強（較弱）熱通量變化，維持較強（較弱）的大尺度環流與降雨變化型態，在此較穩定（較不穩定）的動力與物理機制之下，得以維持臺灣夏季溫度之高（低）可預報度。當臺灣鄰近地區為偏冷變化時，影響臺灣鄰近區域夏季溫度可預報度變化的主因不再是海溫距平強度，而是臺灣鄰近區域是否存在主導該區氣候變化型態的主要海溫距平中心，即所謂的活動中心（center of action），當臺灣東側洋面出現海溫活動中心時，導引熱源變化，激發強盛環流距平籠罩臺灣地區，維持該區氣候變化型態之穩定性，而使可預報度提高；反之，若無主要海溫活動中心存在時，對應較微弱熱通量變化，使臺灣鄰近地區落於環流距平微弱之處，難以維持氣候變化之穩定性，可預報度因此也隨之降低。本文研究結果指出，影響GFS模擬臺灣夏季氣候之准確性與可預報度之主要因素，為臺灣鄰近海域之海溫變化特性，而非與ENSO相關的熱帶太平洋區海溫變化。

　　　　　A 10-member ensemble climate (1950-2000) simulation using the CWB GFS (T42L 18) model was analyzed to study GFS' capability in simulating Taiwan summer (JJA) climate variability, regarding the aspects of simulation accuracy and predictability. The role of SST anomalies in affecting the above model capability is also investigated. Major findings of this study are as follows. Regarding the simulation accuracy, temperature variability in Taiwan is reasonably simulated by the GFS with respect to interannual variability and long-term warming featured by abrupt climate change occurring in the late 1970s. Major systematic error turns out to be the nearly opposite phase of precipitation anomaly in Taiwan. Analysis results indicate that SST anomalies in the oceans surrounding Taiwan provide direct impacts on simulated temperature variability in Taiwan, leading to relatively correct simulation of this temperature variability. The same SST anomalies exert oceanic forcing to regulate model precipitation in Taiwan. However, in observation oceanic variability in this region is regulated by atmospheric forcing. Due to erroneous simulation of the ocean-atmosphere interaction processes, the GFS precipitation anomaly in Taiwan is primarily out of phase with the observed. Regarding the predictability, degree of predictability of Taiwan temperature variability in the simulation is primarily affected by strength of SST anomalies when there are warm anomalies in Taiwan region. Strong (weak) SST anomalies in the vicinity of Taiwan induce strong (weak) surface heat anomalies which tend to maintain strong (weak) large-scale circulation and precipitation anomalies. The above strong (weak) physical and dynamic mechanisms maintain stable (less stable) climate pattern in Taiwan, leading to high (low) predictability in Taiwan temperature variability. Given cold anomalies in Taiwan region, degree of predictability of Taiwan temperature anomalies is mainly influenced by anomalous SST pattern, rather strength of SST anomaly. The presence of dominating SST anomaly behaves as a center of action to induce strong heating variability, resulting in strong and stable circulation anomalies overlying Taiwan to maintain high predictability of temperature variability. On the other hand, the lack of a center of action of SST anomalies surrounding Taiwan corresponds to weak circulation anomalies overlying Taiwan, leading to low predictability in Taiwan temperature variability. In summary, SST anomalies at the vicinity of Taiwan are more important than tropical SST anomalies in affecting the simulation accuracy and predictability of Taiwan summer climate variability in the GFS simulation.