2016 Time-lapse image-domain tomography using adjoint-state methods.pdfVIP

GEOPHYSICS, VOL. 78, NO. 4 (JULY-AUGUST 2013); P. A29–A33, 4 FIGS. 10.1190/GEO2013-0044.1 / g r o . g e s . y r Time-lapse image-domain tomography using adjoint-state methods a r b i l / / : p t t h t 1 2 2 a Jeffrey Shragge , Tongning Yang , and Paul Sava e s U f o s m r complementary image-domain tomography (IDT) strategies (Alber- e T ABSTRACT tin et al., 2006; Yang and Sava, 2011). Rather than posing an ob- e e jective function based on differences between modeled and field s ; t Adjoint-state methods (ASMs) have proven successful for data, the more kinematically biased IDT inversion approaches h g calculating the gradients of the functionals commonly found use objective functions that measure observed imperfections in i r y in geophysical inverse problems. The 3D ASM image- migrated images (i.e., poorly focused extended image gathers p o domain tomography (IDT) formulation of the seismic veloc- [EIGs]). Similar to data-domain FWI approaches, these imper- c r ity estimation problem highlights imperfections in migrated fections are back projected using ASM machinery to form gradient o e image volumes and, using appropriate penalty functions estimates appropriate for velocity model updating. s n e (e.g., differential semblance), forms an objective func