Surface-based model conditioning using an hybrid optimization: methodology and application

Surface-based 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 inferring the right combinations of parameters that match data. Solving such inverse problem is time-consuming and expensive because those models are complex and highly parameterized. In this context, Bertoncello et al. (2010) proposed a new approach to condition surface-based models to data. The key idea behind the method is to decrease the dimensionality of the problem. Three complementary approaches are developed for this purpose. The first approach aims at identifying the leading uncertainty. The second one is a sensitivity analysis on the input parameters. The third one is a re-formulation of the full inverse problem. This re-formulation enables to solve the inverse problem sequentially and not all at once. Each sequence of optimization is solved using the Nelder-Mead simplex approach (Nelder & Mead, 1965). However, this optimization algorithm is only able to handle few parameters at a time, which is a problem for complex real case-studies. In this