Reports: G8
48375-G8 Evaluating Overburden Impacts on Geophysical Responses and Their Applicability in Deep Water Reservoir Characterization Through Inverse and Forward Modeling
Seismic methods have been used extensively for studying subsurface hydrocarbon reservoirs, but the insensitivity of seismic responses to hydrocarbon/water saturations create difficulty when attempting to estimate reservoir fluid saturations. To better characterize the economic yield of a given reservoir, electromagnetic (EM) data can be utilized. As noted by Constable et al. (2007), marine controlled-source EM could be the most important geophysical technique to emerge in many years. The technology offers a great advancement in hydrocarbon exploration due to its sensitivity to the parameters (i.e. porosity, oil/gas saturation) that are associated with hydrocarbon reservoirs (Hou et al., 2005) and has shown potential for hydrocarbon reservoir monitoring (Orange et al., 2009). Seismic and CSEM methods can be used individually or jointly to aid in identifying deep water reservoirs, but overburden parameters (e.g., sediment thickness, seawater depth) vary for different reservoir situations, and play different roles in affecting geophysical responses. The effects of overburden properties can cause misinterpretation of data and erroneous estimation of reservoir hydrocarbon saturations. However, few studies have systematically examined the effect that the overburden has on seismic and CSEM responses.
There have been several studies that consider the overburden’s effect on determining reservoir estimations via seismic methods (Adriansyah and McMechan, 1998; Shen et al., 2002; Angerer et al., 2003; Angelov et al., 2005; Holvik and Amundsen, 2005; Malme et al., 2005; Luo et al., 2005, 2007) that suggest removal of the P-wave overburden effect from the geophysical response could achieve a more accurate reservoir analysis. Most studies look at wave amplitude variations with incidence angle (
Representative baseline models were generated from well log data from field sites in the North Sea, the Gulf of Mexico and Western Pennsylvania. From these baseline models, overburden parameters such as seawater depth, sediment thickness and electrical conductivity, and reservoir parameters (e.g., porosity, gas saturation, reservoir layer thickness) are varied to generate various subsurface models. For each model, geophysical (seismic/EM) forward modeling are used to generate the corresponding geophysical responses. Comparing the responses to the baseline responses we can then determine the sensitivity of the responses of each geophysical method to each parameter or the combination of different parameters.
After the output for both seismic and EM responses are generated, the responses are analyzed and compared. For the seismic method, we study the change in magnitude and EM signals (phases and amplitudes) in response to changes in overburden and reservoir parameters. In order to explore all possibitilies of realistic conditions, we adopt quasi-Monto Carlo method to generated samples from the prior probabilistic density functions, which are derived from prior information (e.g., welllog data) of these parameters, using minimum-relative entropy method (Hou, et al., 2006).
Using data from the forward modeling analysis, we evaluated the relative importance of the overburden and reservoir parameters/attributes applicability on geophysical responses using statistical analyses such as generalized linear model fitting and goodness of fit tests. The analyses give ranks of significance, which enables us to treat certain parameters as deterministic variables. The result helps reduce the dimension of the parameters space, which is critical considering the computing demand for forward model calculations.
Next, an MRE-Bayesian-quasi Monto Carlo inverse modeling is adopted in this study for parameter estimation (Hou et al., 2006). In this study, the framework is improved by introducing an algorithm to calculate intrinsic variances for various data types collected at different times and locations. The variance information is used to normalize the deviations between calculated responses and observations, such that proper weights can be assigned to the posterior samples. Also a subroutine is introduced to deal with the spatial-temporal correlation between the seismic/EM data collected at different locations or seismic angles.
Two graduate students have been trained in programming, statistical analyses, and geophysical data interpretation. For the proposed study, they have made great progress in identifying favorable conditions for seismic/EM techniques, reducing the dimension of parameter space, and developing a robust inverse modeling algorithm, which will be tested using field data. The research results were presented at fall 2008 AGU conference and BAPG symposium. The PI will give another talk on recent advances at AGU in Dec 2009.