Lawrence Berkeley National Laboratory

Abstract We develop a data‐driven clustering method to evaluate a velocity model using surface wave velocity dispersion. This is done by first computing theoretical dispersion curves for 1‐D velocity profiles of all the grid locations and then splitting the resulting dispersion curves into a certain number of groups via the K‐means clustering. The observed dispersion curves are also clustered following the same procedure and the velocity model is assessed by comparing the spatial patterns obtained for the observed and synthetic data sets. The method is applied to evaluate two community velocity models in southern California, CVM‐S4.26 and CVM‐H15.1, using phase velocity maps derived for 3–16 s Rayleigh waves. We found a good correlation in the spatial distribution of clusters between the result of CVM‐S4.26 and that of the observed data, suggesting that the CVM‐S4.26 fits the observed dispersion maps better than the CVM‐H15.1 in terms of features extracted from the clustering analysis. Plain Language Summary With increasing volume of recorded seismic data, various velocity models are often derived for the same region using different data sets and seismic networks with different spatial coverage and resolution. Therefore, evaluating all the existing velocity models in the overlapping region can provide crucial information to future development of tomographic models, such as constructing a standard model by merging all the velocity models. As a machine learning technique, clustering analysis has proven its ability to extract hidden grouping features from large unlabeled data sets. In this study, we develop a simple workflow that utilizes a specific (K‐means) clustering method to evaluate the velocity model. Instead of applying the clustering method directly to the velocity model, we first calculate theoretical predictions for a certain measurable parameter (phase velocity of Rayleigh wave) using the input model and assess the model by comparing the clustering results obtained for the synthetic and observed data sets. The proposed model evaluation method is applied to the well‐maintained community velocity models, CVM‐H15.1 and CVMS‐4.26, in Southern California. The result suggests that CVM‐S4.26 is much better than CVM‐H15.1 for structures in the top ∼20 km. Key Points We develop a data‐driven method that evaluates a velocity model using the K‐means clustering and Rayleigh wave phase velocity dispersion The model evaluation method is applied to community velocity models, CVM‐S4.26 and CVM‐H15.1, in Southern California The result suggests that CVM‐S4.26 gets an evaluation score ∼3 times higher than that of CVM‐H15.1 for structures in the top ∼20 km