Automatic selection of ambient noise observation stations based on the clustering algorithm

Abstract

In order to avoid increasing the workload of correlation function calculation for ambient noise tomography from intensive observation stations, a clustering method based on improved DBSCAN for ambient noise observation stations algorithm is proposed to improve data processing efficiency. According to the ambient noise tomography principle, the main influencing factors of Green's function retrieving are analyzed. Combined with the actual situation of ambient noise observation station arrangement, the selection method of main parameters in cluster algorithm is given. 155 seismic observatory stations in the North America are clustered to improve data processing efficiency. The results show that the overall efficiency of correlation function calculation and superposition is increased by 15.1%, the total time of extraction and screening of dispersion curve is reduced by 18.7%, and the average time of ambient noise tomography data processing is reduced by 12.6% compared with that before clustering, while the quality of ambient noise tomography is guaranteed by clustering processing of intensive ambient noise observation stations.