Joint inversion of first arrival seismic travel-time and gravity data

Abstract

Joint interpretation of disparate geophysical datasets can be a powerful tool in deriving meaningful subsurface geologic models. Here we describe a method for joint inversion of first arrival travel-time and gravity data with application to field data from a geologically complex subduction zone. As both tomographic and gravity inversions are prone to non-uniqueness, incorporation of prior information in the model description is crucial to the success of such algorithms. We employ a layer-based model description, in which interfaces (which may also be called iso-velocity lines) are defined by a summation of arc-tangent functions. Arc-tangent functions are highly flexible in mapping smooth interfaces as well as the nearly discontinuous changes in depth of an interface. Within each layer, the velocity is assumed to vary linearly with depth at each surface location. Because of the non-uniqueness of the gravity inversion, we use prior knowledge to relate the velocity to density values. In our application here, the density is related to the velocity using a fourth-order polynomial whose coefficients are assumed to be known. The nonlinear optimization problem is solved by a very fast simulated annealing (VFSA) technique. At each iteration, travel times are generated by the solution of the Eikonal equation while the gravity anomalies are computed using a standard formula. The objective function consists of two parts: one measures the misfit in travel time and the other measures the misfit of gravity anomalies. We applied our technique to field data collected over the Ryukyu subduction zone offshore Taiwan during an ocean bottom seismometer (OBS) experiment (called TAICRUST) conducted in the year 1995. Application to one NS and one EW trending line resulted in acceptable fit of both travel-time and gravity data. The resulting models are helpful in the interpretation of the local geology.