The description of anisotropy velocities and raypaths as sums of smooth basis was relatively simple to code, and a generic Gauss-Newton optimization algorithm greatly simplified their optimization. The non-uniqueness of the first synthetic inversion is not surprising and will not prevent us from making use of well-determined components of velocities.
The 2D isotropic synthetic converges to the smoothest and most stratified velocity model that will adequately model the picked traveltimes. 3D and anisotropic models could easily increase the non-uniqueness of the inversion. Next, I must test 3D isotropic models to discover what patterns of 3D coverage can isolate velocity anomalies. It would be very useful to know if a typical 3D marine acquisition is adequate, with offsets extending only along one azimuth.
I must test 2D and 3D anisotropic data for non-uniqueness. An isotropic inversion of anisotropically calculated times will demonstrate whether such data can be explained isotropically. With crosswell surveys many have seen that a simple anisotropic model explains the same traveltimes as an isotropic model with much more complicated spatial variations in velocity. Possibly an analogous situation will occur with surface diving waves. I must then make choices on whether to discourage spatial variations or anisotropy more, while allowing for non-uniqueness in both components.