Delivery4D: an open–source model–based Bayesian seismic inversion program for time-lapse problems

An extension of the open–source Bayesian AVO inversion program Delivery [1],[2] to time–lapse seismic inversion problems is described. The inverse problem is for a trace–wise layer–stack model comprising unknown layer–times, rock properties, fluid types and saturations, and pore pressure changes. The “data” are true–amplitude imaged/migrated seismic reflectivity traces, for arbitrary numbers of stack angles and vintages. Both maximum aposteriori models and posterior samples, via MCMC, are made available. Coupling to stress is made via local calibrated regression models, and saturation via Gassmann’s relations. Test problems demonstrate upper limits on what might be reasonably inferred about saturation from seismic data alone, and show distinct refinement in “swept” zones, but small improvements elsewhere.