| Seismic techniques have been widely used for hydrocarbon exploration. Recently, electromagnetic (EM) methods have drawn great attention in the area of reservoir characterization. However, few efforts have been made to explore the reservoir overburden impacts that affect the reliability of seismic, EM or joint methods for exploration. It is essential for successful reservoir hydrocarbon estimation to identify favorable overburden and reservoir conditions where reservoir fluid saturation can be reliably inferred from seismic data, EM data, or integrated geophysical information. This study evaluates the sensitivity of seismic and electromagnetic (EM) responses to hydrocarbon reservoir and overburden attributes through forward and inverse modeling. First I generated baseline truth models from well log data, then varied the overburden and reservoir parameters (e.g., target layer thickness, porosity, saturations, etc.) within their physical ranges, producing multiple realistic subsurface models. Next, geophysical (seismic and EM) responses were computed for each model using seismic/EM forward modeling algorithms, and the differences in seismic and EM data between these realistic and baseline models were analyzed. Forward modeling results suggest that EM calculations are sensitive to several factors both in the reservoir and the overburden. Outside the reservoir, EM responses are most sensitive to seawater electrical conductivity, seawater depth and overburden sediment thickness. While among the parameters/properties in the reservoir, EM responses are most sensitive to gas saturations. Seismic data show strong responses to reservoir parameters, most notably to reservoir porosity. After identifying the significant parameters, regression analyses using generalized linear models were performed and analyzed. Subsequent to exploring the sensitivity of seismic and EM responses to changes in overburden and reservoir parameters/attributes, the applicability of using these datasets for inverse modeling (parameter estimation) was evaluated using a Minimum Relative Entropy-based Bayesian stochastic inversion. Results suggest that estimations of reservoir parameters using high dimensional models can be rather difficult to do with great confidence. It was found that the more prior information available on the overburden unknowns, the closer the posterior estimation of the reservoir parameters will be to the actual values. Results also show that seismic methods are effective in estimating reservoir porosity, while EM has the potential for estimating reservoir saturations. The integration of both, assuming enough information about the priors are known, has the potential to accurately estimate reservoir parameters presuming enough models are run. Informative prior knowledge about the field site, efficient sampling techniques, reduction of the parameter space, and computing capability are all necessary components for successful geophysical characterization of a petroleum reservoir, given the high dimensionality and non-uniqueness of the inverse problem. |
