New features:
- Fixed multiple bugs related to importing variable receiver positions for each shot. This allows seismic forward modelling and FWI e.g. for streamer geometries.
- Removed deprecated N_STREAMER option
- Created Python code PythonIO together with a Jupyter Notebook to create input and model files for DENISE Black-Edition allowing a step-by-step definition of all necessary files for seismic forward modelling and FWI. The Jupyter Notebook also checks if the CFL & grid dispersion criteria are satisfied, a reasonable domain decomposition is defined, if sources and receivers are placed outside of the model or inside the PMLs and some FWI parameters are consistent with the acquisition geometry. An example for the Marmousi-2 test problem can be found here
- Fixed failing AC FWI bug found by Albert Zhang when trying to invert hydrophone (pressure) data
- Added acoustic gradients in velocity-pressure formulation according to Hu (2012)
- Added AGC weighted l2-norm objective function to enhance weak reflection events compared to surface waves in the field data
- Oleg Ovcharenko and Vladimir Kazei added an interactive DENISE Python API which not only allows the definition of input and model files, but also runs the forward/FWI code from within a Jupyter notebook. Furthermore, modelled seismic shot gathers and FWI results can be visualized and analyzed. Using Jupyter notebooks together with the API allows to document the changes of input parameters relative to the standard input files for the Marmousi-2 benchmark problem and creates a new set of input files without changing the original files. Therefore, foward modelling and FWI runs can be very effectively documented in a reproducible way. An example for the Marmousi-2 problem can be found here
- Multiple smaller bug fixes