First Break - Volume 41, Issue 9, 2023

Distributed acoustic sensing (DAS) data recording a large amount of the subsurface information become more promising for real-time seismic monitoring. Therefore, fast and accurate imaging techniques are required to handle large datasets. Besides the issue of the computation cost, most of the migration methods, such as reverse-time migration (RTM) and Kirchhoff migration suffer from the artifacts, influencing the quality of the image, because of the limited-aperture data. Here, we perform a migration, based upon the Fresnel volume on the simulated geophone VSP and DAS VSP acquired by newly developed fibre-optic cables in Canada. We show that the Fresnel volume migration with a competitive runtime is superior and robust compared to the RTM and Kirchhoff migration. The angle-domain common-image gathers (ADCIGs) extracted from the Fresnel volume migration is more reliable and cleaner than that of the conventional Kirchhoff migration, used further for the AVO analysis and inversion.

For too long, reservoir modelling platforms have encouraged users to invest substantial resources into building most likely (‘base case’) representations of each component of the overall subsurface, including the structure, facies, petrophysics, and fluid model. These representations are aggregated into a ‘best technical case’ reservoir model. However, each time new data is available, this best technical case model must be updated or even rebuilt. Also, later in the reservoir life, major discrepancies are often observed between the dynamic data (e.g., pressure or production) and the model predictions. Approaches traditionally used to align the model with the reservoir history focus on certain model components (e.g., porosity edited using multipliers) at the expense of other important components (e.g., structure). Ignoring some model components hinders the accuracy of predictions, and the model loses its usefulness as a decision-making tool.

Our solution is to treat reservoir modelling as an inverse problem. The reservoir data provides insight into the unknown reservoir properties, but with large uncertainties. To account for all uncertainties, we present a new probabilistic approach, Unified Ensemble Modelling: an ensemble of equiprobable models is built, representing the unknown reservoir properties through probability distributions, which allows us to assess project risks and optimise reservoir development decisions.

Matching pursuit is a well-known algorithm that enables the decomposition of seismic signal with high temporal and frequency resolutions by adaptively extracting wavelets that locally match the signal. However, this method processes each trace independently and does not take into consideration the lateral continuity of the seismic data. In addition, this approach is very sensitive to subtle changes in the seismic signals leading to non-unique solutions. Furthermore, matching pursuit suffers from performance limitations. In this article, a new method is proposed in which geological lateral continuity is captured using a Relative Geological Time (RGT) model. This technique is a RGT-based multichannel matching pursuit which consists of extracting wavelets using the classical matching pursuit method and laterally propagating the extracted wavelets by following their respective RGT age. The method has been applied to the MAUI dataset, offshore New-Zealand. The results indicate that this method improves the lateral continuity while still respecting the geology. The computing time required to extract the iso-frequencies is tremendously lessened, and the extracted iso-frequencies are less sensitive to random noise.

The results of a seismic inversion method designed to screen for AVA anomalies in siliciclastic frontier basins is tested in a mature deep-water setting, the Mississippi Canyon, Gulf of Mexico. The method, which is founded on a universal rock property model for siliciclastic sediments, uses widely available partial stacks to invert seismic data for intercept and gradient impedances and generate a relative elastic inversion volume (rEEI), optimised for lithology and fluid detection. The results demonstrate the method is highly effective, but not infallible, at identifying hydrocarbons without any well data control in the simply buried Neogene and Paleogene (Tertiary) strata of the Mississippi Canyon. They further suggest that the use of the method in the early stages of prospecting in simply buried, siliciclastic basins globally, has the potential to identify opportunities that might otherwise be overlooked when interpreting only on the full stack.