An encoder-decoder deep surrogate for reverse time migration in seismic imaging under uncertainty

Seismic imaging faces challenges due to the presence of several uncertainty sources. Uncertainties exist in data measurements, source positioning, and subsurface geophysical properties. Reverse time migration (RTM) is a high-resolution depth approach useful for extracting information such as reservoir localization boundaries. RTM, however, time-consuming data-intensive it requires computing twice wave equation generate store an condition. when embedded quantification algorithm (like Monte Carlo method), shows many-fold increase its computational complexity high input-output dimensionality. In this work, we propose encoder-decoder deep learning surrogate model RTM under uncertainty. Inputs are ensemble velocity fields, expressing uncertainty, outputs seismic images. We show by numerical experimentation that can reproduce images accurately, and, more importantly, propagation from input fields image ensemble.