We have demonstrated a target-oriented joint inversion method, based an iterative least-squares migration/inversion, for imaging incomplete time-lapse seismic data sets.
By posing time-lapse imaging as a joint inversion problem, the RJMI method attenuates uncorrected artifacts caused by gaps in the monitor acquisition geometries.
We considered an undershoot problem, where obstructions prevent perfect repetition of acquisition geometries for different surveys and a sparse time-lapse data problem, where a random fraction of the monitor data sets are recorded.
In both numerical examples, we showed that joint inversion (within the RJMI framework) produces time-lapse images of the best quality relative to migration, normalization with the Hessian diagonal and separate inversion.
We recognize that both the separate and joint inversion results can be improved with stronger spatial regularization, but it is arguable that such an approach will introduce too much unjustifiable bias into the inversion.
Significant progress made in the field of compressive imaging (Candes and Romberg, 2007) provides a possible pathway for better image recovery from incomplete time-lapse seismic data.
Target-oriented joint inversion of incomplete time-lapse seismic data sets
![[pdf]](icons/pdf.png){kind=icon}