Sharif University of Technology Sharif Journal of Civil Engineering 2676-4768 vol. 25 51.1 2009 12 22 PRINCIPAL COMPONENT ANALYSIS APPLIED TO SEISMIC HORIZON INTERPRETATIONS PRINCIPAL COMPONENT ANALYSIS APPLIED TO SEISMIC HORIZON INTERPRETATIONS 69 76 36 FA H. Sabeti Mineral Group Birjand University of Technology A. Javaherian Dept. of Petroleum Engneerin Amirkabir University of Technology B. N. Araabi Dept. of Electrical and Computer Engineering University of Tehran Journal Article 2007 08 14 One of the most important stages in seismic interpretation is picking especial horizons in order to detect their underground downward and upward movements in an oilfield. Background noise, however, causes many dif- ficulties to this end. Considering a narrow window of a seismic section, whose re ectors are nearly horizontal, and applying a multivariate statistical method called the Principal Component Analysis, we find the largest eigenvalue that has the most contribution to the variance of data. Lower eigenvalues are subject to noise. Projecting data onto an eigenvector associated with the largest eigenvalue, we obtain a trace with sharper peaks and troughs. This method is applied to two synthetic models; horizontal re ectors and anticline. We, also, examine the window length and dominant frequency of the seismic wavelet. Obtained trace with significantly attenuated noise can be used for tracking weak horizons in a seismic section with a signal-to-noise ratio of 0.2. Dominant frequency cannot change the result considerably. Optimum window length is the area in which re ectors are horizontal. It is also applied to the real data of an oilfield in S.W. Iran. The obtained results were useful in picking some important horizons. One of the most important stages in seismic interpretation is picking especial horizons in order to detect their underground downward and upward movements in an oilfield. Background noise, however, causes many dif- ficulties to this end. Considering a narrow window of a seismic section, whose re ectors are nearly horizontal, and applying a multivariate statistical method called the Principal Component Analysis, we find the largest eigenvalue that has the most contribution to the variance of data. Lower eigenvalues are subject to noise. Projecting data onto an eigenvector associated with the largest eigenvalue, we obtain a trace with sharper peaks and troughs. This method is applied to two synthetic models; horizontal re ectors and anticline. We, also, examine the window length and dominant frequency of the seismic wavelet. Obtained trace with significantly attenuated noise can be used for tracking weak horizons in a seismic section with a signal-to-noise ratio of 0.2. Dominant frequency cannot change the result considerably. Optimum window length is the area in which re ectors are horizontal. It is also applied to the real data of an oilfield in S.W. Iran. The obtained results were useful in picking some important horizons. https://sjce.journals.sharif.edu/article_36_e2db1e64bc46f27bb1f50e6cb621bd35.pdf