Geophys. J. zyxwvutsrqpon Int. zyxwvutsrqponml (1993) 114,405-410 zyxwvutsrq RESEARCH NOTE Testing the reliability of Vp/ Vs anomalies in traveltime tomography Albert0 Michelini CNR-lstituto Talassografico Sperirnentale, Viale Romolo Gessi 2, 34123 Trieste, Italy Accepted 1992 December 21. Received 1992 December 18; in original form 1992 February 24 SUMMARY In traveltime tomography, zyxwvu V,/V, models calculated as ratio between the resolved P and zyxwvut S models are affected by the different ray coverages and phase-onset reading accuracies of P and S waves. The result of this is that the computed V,/V, models can display large fluctuations which are unrelated to the true structure. Introduction of some P-S coupling (e.g. proportionality, correlation) as a priori data permits us to stabilize the resulting V,/V, model about a preselected average value and to minimize the insurgence of fictitious V,/V, anomalies. In general, addition of P-S coupling trades off with data misfit and adoption of this technique is suggested for robustness tests of resolved V,/V, features. The technique is applied to a synthetic data set generated using an idealized fault structure and the source-receiver geometry existing at Parkfield, California. Key words: fault zone, seismic tomography, V,/V, ratio. INTRODUCTION In recent years, traveltime tomography has been widely applied to both exploration and seismological seismic data. Initial applications of the method were restricted to P-velocity (V,) structures and extension to S models (Vy) was delayed until more accurate S data became available. However, determinations of zyxwvuts S arrival times are generally difficult because these onsets lie in the P coda with the result that the S-wave data sets available for tomography are often conspicuously smaller and with larger reading errors than those of the associated P-waves. It follows that the S ray coverage is generally more sparse and the determined V, and Vs models may locally diverge with the generation of fictitious V,/Vs anomalies. This is undesirable because determination of V,/V, ratios permits more reliable identification of lithology and/or localization of zones having anomalously high pore-fluid pressures. In this note, I present a technique that allows determination of conditioned Vp/Vs models. This technique is based on the introduction of some variable degree of correlation between V, and V, which, in the formulation of Tarantola zyxwvutsrqp & Valette (1982), corresponds to the use of an a priori non-diagonal model-parameter covariance matrix. Similarly, it corresponds to the use of an a priori datum (Jackson 1979) which establishes some additional coupling or proportionality between V, and V7, This approach was also previously suggested by Mora (1987) in his description of the inversion of common midpoint gathers in seismic exploration, and by Kennett, Sambridge & Williamson (1988) in their scheme for multiparameter classes inversion. In this note, the technique is applied to a synthetic data set consisting mainly of earthquake arrival times which was generated using the source-receiver geometry of the Parkfield, California, seismic network. The true model used to generate the arrival times consists of a low-velocity fault-zone structure. In fact, motivation for the develop- ment of the technique came from the results of Michelini & McEvilly (1991) who found an anomalously high value of V,,/V, in the nucleation of the 1966 Parkfield earthquake. METHOD Because traveltime tomography is well established, only the relevant equations needed to implement existing computer algorithms that solve for both V, and V, models are presented here. When the data set consists of earthquake arrival times, I assume that the determination of the hypocentre parameters has been already carried out (e.g. application of the separation of parameters method; Pavlis & Booker 1980, among others). In general, we can relate the traveltime residuals, atP and z 86, determined as the difference between the observed and calculated times with some first-order perturbations in the V, and V, model parameters, amP and amS (1) 405 by guest on July 27, 2016 http://gji.oxfordjournals.org/ Downloaded from