現今的天氣預報，十分仰賴數值天氣模式的資訊，因此，為了評估其可信度，必須對模式進行精確性分析。隨著衛星資料應用的與時俱進，其高時空解析度以及不受天候影響的優點，近年來成為驗證模式的工具之一。本研究即使用衛星資料，針對三個全球預報系統(Central Weather Bureau Global Forecast System；CWB GFS, National Center for Environmental Prediction Final Operational Global Analysis data；NCEP FNL, European Center for Medium Range Weather Forecasting (ECMWF) ERA-interim)之模擬衛星觀測，以及針對CWB GFS的雲參數特性，利用實際衛星資料進行評析。 於評估的時間與空間區間中，首先進行亮度溫度評估，此部分工作使用Community Radiative Transfer Model (CRTM)將模式輸出之地球物理量，正演成衛星所觀測到的亮溫，並與地球同步衛星所觀測到的亮溫進行分析。初步結果顯示，CWB GFS與NECP FNL表現非常相似，但高層薄雲的模擬可能較少或低估，而ERA-interim則在模擬雲頂溫度有最低溫約220K的閾值限制。 另外一方面，也進行了雲參數(即Cloud Fraction (CF)、Cloud Top Pressure (CTP)及Cloud Optical Thickness(COT))的特性評估，使用Cloud Feedback Model Intercomparison Project (CFMIP) Observation Simulator Package (COSP)，進行模式與MODIS雲產品比較。分析結果顯示CWB GFS的高雲、薄雲太少，中層雲出現頻率過高且低雲有較大的低估，在非洲及澳洲沙漠區域的雲模擬次數過高，整體而言原因來自厚雲出現頻率過高、薄雲低估。評析總雲量平均的結果發現，沿岸處雲量高估、南海地區雲量高估，ITCZ (Intertropical Convergence Zone)下沉區雲量較為低估。 ;The accurate cloud information is critical in the NWP (Numerical Weather Prediction) models, which might associate with improved forecast skills. Nowadays, meteorological satellites provide wide spatial coverage with multi channels’ spectrum of reflectance and/or radiances. Satellite-retrieved products are assumed to be a useful proxy data sets to evaluate the NWP model’s performance other than geophysical parameters. In this study, the evaluation of clouds in NWP models based on comparisons of observed and simulated climatologies of BT (Brightness temperature) and cloud fraction. We adopt model analysis field during in the whole month of January in 2011. We would like to firstly use fast Community Radiative Transfer Model (CRTM) to evaluate Central Weather Bureau Global Forecast System (CWB GFS), National Center for Environmental Prediction Final Operational Global Analysis data (NCEP FNL), European Center for Medium Range Weather Forecasting (ERA-interim) model from calculated brightness temperature against to MTSAT-2. The initial results show that CWB GFS and NECP FNL perform very similar and both of them might underestimate the high level cloud. ERA-interim has threshold at the 220K region. We secondly use Cloud Feedback Model Intercomparison Project (CFMIP) Observation Simulator Package (COSP) to evaluate CWB GFS cloud properties, and use the simulated cloud features to compare with MODIS level 3 cloud products. The results show that total cloud fraction perform well, but some overestimate at coastal region, some underestimate at ITCZ region. For cloud occurrence analysis, the results show some overestimate in Africa and Australia. Therefore, we try to combine the cloud top pressure and cloud optical thickness for further analysis. The overall results show model has various degree of error. These errors include an underestimation of optically thin clouds, high level cloud and low-level cloud; an overestimation clouds in middle levels and thick cloud. The reason might be caused by the cumulus parameterization scheme or ice water content scheme and so on.
