Wang et al. ambient noise adjoint tomography (ANAT) workflow
bch0w opened this issue · 4 comments
My current research focuses on ambient noise adjoint tomography which follows a similar workflow as earthquake inversions, however the scheduling and processing of simulation results differ pretty significantly.
For example, simulations are provided in force, and run once for each orthogonal direction (E, N, Z) for both forward and adjoint simulations. Similarly rather than having events, each station in the stations file will act as a 'source' station. Other significant variations will need to be accounted for throughout the package.
SeisFlows is ready to be tuned for such a workflow, and I envision anat_fwd
and anat_inv
(or something like that) workflows that will automatically simulate FORCESOLUTION files for each orthogonal direction, with new (or re-tooled) preprocessing modules performing the appropriate SGF pre-processing as defined by Wang et al. 2019
@evcano I see you have a noise_workflow branch, just wondering if that is related to this topic?
Hi @bch0w,
Yes, it is related. I am working on a new type of workflow for SeisFlows to conduct FWI using noise correlations. However, the approach I use is quite different from that Wang et al. 2019. My workflow follows the approach of Tromp et al. 2010. Instead of modeling Green's functions, I compute ensemble noise correlations and sensitivity kernels using the noise simulations implemented in SPECFEM.
I was planing to talk to you about it once I finished, are you interested?