Automatic estimation of traveltime parameters in VTI media using similarity-weighted clustering

Compared with hyperbolic velocity estimation methods, nonhyperbolic methods (such as shifted hyperbola) are better choices for large offsets or vertical transverse isotropy (VTI) media. Since local seismic event slope contains subsurface information, they can be used to estimate zero-offset two-way traveltime and normal moveout velocity. The traditional velocity estimation methods require a great deal of manual work and are also prone to human error. In order to estimate the traveltime parameters for VTI media automatically, in this paper, we propose to use predictive painting and similarity-weighted clustering to obtain traveltime parameters. The predictive painting is used to estimate zero-offset two-way traveltime, and the shifted-hyperbola traveltime equation is used to obtain velocity and anisotropy attributes. We first map local slopes to zero-offset two-way traveltime and moveout-parameters domain and then use similarity-weighted k-means clustering to find the maximum likelihood anisotropy parameters of the main subsurface structures. In order to demonstrate that, we apply the similarity-weighted clustering method to synthetic and field data examples and the results are of higher accuracy when compared to the ones obtained using multiparameter semblance-based method. From estimation error section, it can be seen that the estimation error of multiparameter semblance-based method is about 3–5 times that of the proposed method.