Part 4 #GigatonneCCS

In Part 4 of this series, we demonstrate how we conducted more detailed investigations of the Elephant site and explored how specific subsurface evaluation methods are particularly valuable and can help derisk some of the critical elements of a deep saline aquifer – lithology typing and distribution. We apply quantitative seismic interpretation techniques on high-quality broadband 3D seismic data, and in this installment describe in more detail how this was done to characterize and derisk the Elephant concept.

Previously in Parts 1-3

In previous articles on the Elephant site, located within the PGS19MO2NWS 3D GeoStreamer dataset in the Norwegian Sea, we explored defining the concept, the interpretation, and geological elements of the site – a large, migration-assisted aquifer opportunity within Jurassic sandstone aquifers deposited along the Norwegian margin. 

Quantitative Seismic Interpretation | A Key Tool For Derisking

Experience has shown that injectivity can be a key risk to the successful operation of a carbon storage site. Although this can be for operational reasons, aquifer properties are also key risk area in subsurface site characterization. Consequently, an important requirement in assessing site suitability for injection and considering potential pressure management risks in saline aquifers, is to characterize the quality and extent of the sandstone in the storage targets. Identifying the presence and extent of porous units within the seal and overburden units is also highly valuable when considering the potential for seal bypass and overall containment risk assessment. Finally, an effective definition of the architecture and extent of the units also allows effective capacity estimates to be made.

A specific challenge in saline aquifer targets, particularly those that have not been extensively drilled during E&P activities, is the uncertainty in aquifer properties. Methods that can help directly characterize the presence and extent of these units, and give some indications of potential quality, are highly useful. This is particularly the case in areas adjacent to producing regions, where there is a low density of well data to calibrate estimates and models.

This article shows how seismic reservoir characterization was used at the Elephant site to offset low well density and provide critical insights into the site’s characteristics.

Describing the Method

Seismic Reservoir Characterization requires a multi-disciplinary approach, bringing together quantitative geophysics, petrophysics and geology. However, the key to achieving accurate results is assessing up-front whether the data is fit for the task. The initial step requires pre-conditioning of the seismic data, which involves addressing acquisition and processing limitations while enhancing the wide bandwidth of the resultant seismic signal. The overall aim is to improve the Signal to Noise Ratio (SNR), ensuring that meaningful features stand out and that the AVO (Amplitude versus Offset) is preserved and enhanced.

Another key step is to establish the link between the seismic and the well data and calibrate seismically derived properties with information from well logs and key petrophysical properties of interest. This process forms the foundation of rock physics modeling.  Having established a link between reservoir and elastic properties we can now develop insights into the underlying geology within the seismic area of interest, allowing us to use the outputs to condition geological and reservoir models directly derived from the data. And because the data is 3D this ability to assess lithology characteristics within the volume means we can directly connect modeled properties in the reservoir model to a measurement from the subsurface, rather than using pure geostatistical methods to interpolate and ‘fill in the gaps’ in knowledge of aquifer distribution and connectivity.

The optimized seismic dataset is subsequently inverted either for the elastic properties, which can be obtained directly from the AVO information (P-Impedance and S-Impedance using simultaneous seismic inversion) and interpreted with rock physics modeling as a guide or within one of the numerous inversion workflows utilizing rock physics models and yielding a statistical representation of petrophysical parameters directly. 

Getting Into The Detail

Our investigations of The Elephant dataset (PGS19MO2NWS), followed this general workflow:

• Reservoir Oriented Processing (ResOP) or seismic data preconditioning
• Rock physics analysis and modeling
• Facies inversion: Combining the findings from both steps to create a prediction of lithology type.

In the preconditioning stage (ResOP), we applied spectral shaping to enhance temporal resolution while maximizing the wavelet peak to the sidelobe ratio. Spectral matching between the partial stacks was followed with residual trim statics alignment. At the end of the ResOP, seismic data is characterized by broad spectrum, high SNR, and flat primary events within the gather – which is a prerequisite for successful inversion work.

Building rock physics models to calibrate seismic reservoir characterization predictions for the Elephant site presented a significant challenge, primarily due to the scarcity of well control. The Trøndelag Platform, historically not regarded as highly prospective for hydrocarbons, offers an unique advantage for CO2 storage projects - minimal co-location issues with oil and gas-related activities. However, this advantage comes with the challenge of low well density which would have been highly useful for site characterization.

To overcome this limitation, we leveraged our comprehensive regional Norwegian Sea petrophysics and rock physics database (rockAVO). Additionally, we augmented this database with petrophysical analyses from five additional proximal wells. The goal was to construct a generalized regional understanding of petrophysical properties within Jurassic aquifers.

In this context, our reliance on high-quality seismic data becomes even more pronounced – which the Elephant dataset provides. The well-to-seismic tie of 6508/5-1 (a key offset well in the vicinity of our area of interest) is quantified by the cross-correlation coefficient exceeding 0.81 within the Jurassic/Triassic 1-second two-way travel time (TWT) window post preconditioning and this underscores the dataset’s suitability for this method.

In our pursuit of robust seismic reservoir characterization, we investigated four separate lithofacies models - a critical input for our 1-step facies inversion workflow. These models serve as depth-dependent representations of the elastic properties inherent to specific rock units.

1. Hot shales: This litho-fluid class gathers the properties of the hot shales found in the Spekk Formation. Notably, the top Melke Fm. shales share striking similarities with their younger Spekk counterparts, while deeper within the formation, they gradually assume elastic properties akin to typical Jurassic shales. This is important for mapping the main seal to the CO2 storage site.
2. Generic shales of deeper Jurassic intervals: These shales, spanning deeper Jurassic layers, pose challenges due to their acoustic properties' remarkable similarity to Jurassic sandstones.
3. Jurassic sandstones: The acoustic resemblance between Jurassic sandstones and shales underscores the complexity of interpretation. Within seismic data, we observe subtle reflectivities within the Bat and Fangst groups, further complicating the delineation of formations like Top Tilje or Top Ile. The separation between these formations lies solely within the elastic domain, specifically the Vp/Vs ratios - an interesting challenge for the inversion methodology.
4. Carbonates: Occasionally present within Bat and Fangst Gp., their fast velocities would tend to be misinterpreted by facies classification – which warranted their separate litho-fluid class.

Unlike traditional simultaneous inversion methods, which involve perturbing data in P-Impedance/S-Impedance space, followed by seismic modeling, comparison with angle stacks, and iterative lithology-fluid class model updates, our 1-step facies inversion takes a more direct route to estimate lithology classes (and fluid, although not in this case).

Here is how it works:

1. Initial model: We begin with a priori stochastic model of lithofacies distribution. This serves as our starting point.
2. Calculating elastic properties and seismic responses: Our approach directly computes elastic properties and corresponding seismic responses from the litho-facies distribution. By integrating rock physics principles, we bridge the gap between lithofacies and seismic data.
3. Iterative refinement: If a mismatch between the modeled and acquired seismic data arises, we update the litho-facies distribution directly.

The outputs of the 1-step lithofacies inversion include most likely facies distribution, as well as a posterior probability distribution of every modeled litho-facies class and the average elastic properties. The key point of this process is to use the seismic to directly connect reservoir property models to the data and offset the absence of well data directly on the site, as best as possible.

The sandstone distribution and connectivity characteristics play a pivotal role in defining the sites overall storage potential. Understanding this distribution helps us delineate potential migration pathways for CO2 plumes and can be directly used to condition simulation models in subsequent steps – these will be the subject of a future article. Encouragingly, we observe possible connections between the Ile Formation and Garn Formation aquifers. Without the high-quality 3D seismic data to identify the ‘plumbing’ between the two formations, we might otherwise have assumed a much simpler stratigraphic model, with no connectivity and communication between the two. These observations could have profoundly different consequences for how a site is modeled and developed.

In addition to the most probable facies distribution, the elastic properties of the subsurface - specifically P-Impedance, S-Impedance, Vp/Vs Ratio, and Density - yield valuable insights. Notably, P-Impedance exhibits a strong correlation with total porosity in clastic formations, particularly where pore fluid variability is minimal, as is expected to be the case in an initially brine-saturated formation.

This transformation from P-Impedance to total porosity informed our static and dynamic reservoir models. In the next installment we will explore how the subsurface analysis, including the work described here, were integrated into reservoir models and the resultant scenarios simulated to examine the storage potential and migration characteristics of the overall system. The complete narrative behind those will unfold in our upcoming installment. Stay tuned!