Abstract
Choosing the right parameterization to describe a transversely isotropic medium with a vertical symmetry axis (VTI) allows us to match the scattering potential of these parameters to the available data in a way that avoids a potential tradeoff and focuses on the parameters to which the data are sensitive. For 2D elastic full-waveform inversion in VTI media of pressure components and for data with a reasonable range of offsets (as with those found in conventional streamer data acquisition systems), assuming that we have a kinematically accurate normal moveout velocity (v(NMO)) and anellipticity parameter eta (or horizontal velocity v(h)) obtained from tomographic methods, a parameterization in terms of horizontal velocity v(h), eta, and epsilon is preferred to the more conventional parameterization in terms of v(v), delta, and epsilon. In the v(h), eta, and epsilon parameterization and for reasonable scattering angles (< 60 degrees), epsilon acts as a "garbage collector" and absorbs most of the amplitude discrepancies between the modeled and observed data, more so when density rho and S-wave velocity V-S are not inverted for (a standard practice with streamer data). On the contrary, in the v(v), delta, and epsilon parameterization,. is mostly sensitive to large scattering angles, leaving v(v) exposed to strong leakages from. mainly. These assertions will be demonstrated on the syntheticMarmousi II as well as a North Sea ocean bottom cable data set, in which inverting for the horizontal velocity rather than the vertical velocity yields more accurate models and migrated images.