Supervision of seismic velocity models of the Reykjanes Peninsula Rift, SW Iceland

Abstract. Most methods for processing seismological data require a suitable velocity model characteristic for the given region being defined. This is also the case of the Reykjanes Peninsula (RP) in SW Iceland, where the REYKJANET seismic network was built to monitor local seismicity in the rift zone. At present, four previously published 1D velocity models (SIL, BRA, TRY and VOG) can potentially be used, prompting us to determine which one is the best. In order to address this issue, we arranged a contest in which all four 1D models and one additional 3D model (T3D) were entered. Uniform methodology for classifying the models was applied and included an analysis of: (i) post-localization travel-time residuals, (ii) residuals of the P-wave first-motion incidence angle and (iii) model-predicted and measured Rayleigh-wave dispersion. We discovered that no single model was unequivocally the most optimal, as the differences between them proved rather minor. A common shortcoming of all the models is the bias of the P-wave first motion incidence angle residuals, which may be a general problem for methods working with P-wave amplitudes (e.g., moment tensor solutions). The VOG model was selected with a weak preference.

Finally, we propose a simple method for modifying any of the 1D models by adding a station-dependent surface layer with a vertical velocity gradient. This way, a pseudo-3D model is generated which is fully competitive with a true 3D model while retaining the simplicity of 1D ray tracing. The efficiency of this correction was demonstrated using the VOG model. The corrected VOG model provides post-localization residuals comparable with the true 3D model T3D, has zero bias in predicting the P-wave first-motion incidence angles, and agrees acceptably in predicting the Rayleigh-wave phase-velocity known from other sources. While calculations with a 3D model can be clumsy, the proposed pseudo-3D model is defined by few parameters and is very easy to use. Its applicability is limited to earthquake sources deeper than the deepest lower limit of the topmost layer below the stations.