Data driven auto-picking of surface wave dispersion curve

Surface waves (SW) are considered as noise in exploration records. However, they are sensitive to the near-surface properties, and can be used to retrieve shear wave velocity (VS) models. VS models have various applications in different fields. For instance, they are used in geophysical surveys for subsurface exploration. Surface wave analysis has been extensively utilized to derive the VS profile. Extraction of dispersion curves (DCs) is an essential step in surface wave analysis. The DCs are typically picked manually on the dispersion image, which can be a time-consuming process and subject to the expertise of the user. In this study, we presented a comprehensive and fully automated approach for extracting dispersion curves. This method eliminates the need for manual intervention or human input throughout the process. To evaluate the reliability and accuracy of the automatically picked dispersion curves, we conducted a comparative analysis with their corresponding dispersion curves that were independently and manually selected. The application of our proposed method was specifically implemented on seismic data obtained from Kefalonia, Greece. This data served as the basis for assessing the performance of our automated technique. Our findings demonstrated a remarkable level of agreement between the automatically picked DCs by our method and those manually picked independently. To quantify the dissimilarity between the two sets of dispersion curves, we calculated the normalized total misfit, a metric that represents the overall discrepancy. The results revealed that the normalized total misfit between the automatically picked DCs and the manually picked ones was determined to be 5.45%. This indicates a relatively small discrepancy and suggests a high degree of accuracy in the automated picking process. One notable advantage of our automated approach is the significant reduction in time required for picking dispersion curves. While manual picking typically demands several hours to complete, our method achieved the same task within a matter of minutes. This substantial reduction in processing time highlights the efficiency and time-saving potential of our automated technique. To summarize, our study has presented a fully automated method for extracting dispersion curves. By conducting a comparative analysis with manually picked dispersion curves, we have confirmed the quality and accuracy of the automated picking process. The seismic data from Kefalonia, Greece served as a valuable test case, demonstrating the strong agreement between the automatically picked dispersion curves and the manually picked ones, with a normalized total misfit of 5.45%. Moreover, the considerable reduction in processing time showcases the efficiency and practicality of our automated approach.