藉由濾波方式可擷取出大氣數據中之10~20天與30~60天季內震盪，為影響颱風運動的重要背景環流。Morlet Wavelet Transform可用來濾波，但是當歷史資料不夠長時，如128天訊號，所產生的邊界效應會使30~60天頻段可信度降低；而直接用Empirical Mode Deconposition(EMD)濾波，則會有隨機的mode-mixing現象產生，造成分析上的不便。本研究先用Morlet Wavelet Transform預處理，再以EMD濾波，可有效率地萃取出10~20天與30~60天波動。 濾波結果顯示：2009年莫拉克颱風侵台前後，風場與外逸長波輻射(OLR)場10~20天與30~60天波動皆為有利引進西南氣流的OLR低相位與西南風，交互作用後，使得與強度相同且路徑相似的2008年鳳凰颱風相比之下，降雨期較長。 從OLR的時間延遲相關分析來看，有利10~20天頻段移動的路徑，大致上與前人研究一致，所以綜合Wavelet和EMD的確可以自即時預報資料128天訊號中萃取出10~20天與30~60天波動。 In this project we extract the 10~20 day and 30~60 day introseasonal oscillations from he atmospheric data and then study and analyze these signals. One way to do that is by Wavelet Transform, and the other is by the method of Empirical Mode Decomposition (EMD). However, there are shortcomings in these methods when each is used alone. In our work we use Morlet Wavelet Transform as a pre-filter to filter out the high frequency signal from the data and then use EMD to further filter the remaing signals. As a result we are able to obtain the desired signals with contentment. Our results show that in the period when typhoon Morakot invaded Taiwan in 2009, the 10~20 day and 30~60 day oscillations were in phase with each other which in turn enhanced Southwest monsoon circulation and moisture transportation and caused more rainfall in Southern Taiwan than that by typhoon Fung-wong in 2008. Using the time-lagged correlation analysis of the Outgoing Longwave Radiation (OLR), we show that the path along which the 10~20 day signal traveled is similar to that shown by others. This implies that the filter method we developed in this study can successfully extract the intraseasonal oscillations.
