Abstract:A ccording to the recently developed sparsesignal theory and the morphological differences be-tween reflections and ground-roll,the shot gatherwhich contains ground-roll is reshaped to be a one-dimensional array,and the morphological compo-nent analysis(MCA) technique is used to separatethe ground-roll.The key of the MCA method is tochoose the appropriate dictionaries.In view of themorphological differences between reflections andground-roll,two kinds of dictionaries are chosen.They are the one-dimensional UDWT dictionaryand the one-dimensional LDCT dictionary.Thefirst one represents sparsely the ground-roll com-ponent in common-shot records,and the secondone represents sparsely the reflections in common-shot records.Block-coordinate relaxation(BCR)algorithm is used to calculate the objective func-tion,and the shot record is decomposed intoground-roll component and reflection component.Then the ground-roll is removed.Real data exam-ples show that the MCA method not only can suc-cessfullv remove ground-roll,but also preserve re-Elections,especially for low frequency and lowwave-number information.