Conditioning a hybrid geostatistical model to wells and seismic data

Hybrid geostatistical models imitate a sequence of depositional events in time. By considering sedimentation processes, these algorithms produce highly realistic subsurface structures from a variety of environments. However, since depositional events are forward-modeled, they cannot be directly conditioned to data. Therefore, conditioning requires solving a possibly expensive inverse problem. In this study, an optimization scheme is developed that allows conditioning turbidite simulation to thickness information and boreholes data. The methodology is based on the addition of a noise to the lobes' surface, which is perturbed until data are satisfactorily matched. The process is made more efficient by dividing the optimization problem into similar steps, allowing decreasing the number of parameters to consider at a time. This methodology is applied to a realistic dataset to demonstrate the validity of the method.