Introduction

Reservoir modeling has long served as the foundational pillar of petroleum exploration and production. For decades, geoscientists and reservoir engineers have meticulously constructed full-field static and dynamic models to forecast hydrocarbon volumes, optimize production, and strategically plan field development. Yet, despite the sophistication of these workflows, a critical driver of reservoir behavior—small-scale, sub-cell heterogeneity—is frequently overlooked. Ignoring these small-scale sedimentary heterogeneities can be a critical error, leading to significant inaccuracies in production forecasting, remaining reserve estimation, and enhanced recovery planning. Overlooking these crucial details not only erodes technical credibility but also results in millions of dollars in poor field development decisions.

This is where SBED (Sedimentary Bedding-scale Heterogeneity Modeling Software), developed by SubsurfaceAI (formerly Geomodeling Technology Corp.), has fundamentally changed the approach to reservoir characterization. SBED pioneered a practical, accessible workflow for integrating sub-cell heterogeneity into mainstream reservoir models. It acts as the essential link between the core plug scale and the simulation grid scale, enabling interpreters and engineers to capture the subtle—yet decisive—effects of bedding-scale variability.

In this comprehensive article, we will delve into the profound reasons why sub-cell heterogeneity is so crucial, examine the costly consequences of ignoring it, and demonstrate how SBED enables its seamless integration into full-field workflows. The insights presented are supported by decades of peer-reviewed research and case studies that validate the immense value of this modeling approach.

What is Sub-Cell Heterogeneity?

In reservoir modeling, we typically work with grids—three-dimensional blocks that represent cubic volumes of rock. These cells are assigned bulk petrophysical properties (such as porosity, permeability, and saturation) derived from seismic data, well logs, and core data. However, the resolution of these grids is inherently limited. Typical simulation grid cells measure 10–50 meters laterally and 1–5 meters vertically. Conversely, sedimentary features like laminations, cross-bedding, inclined heterolithic stratification (IHS), or thin shale drapes often occur at scales of millimeters to centimeters.

This significant mismatch in scale creates a critical blind spot. When these fine-scale structures fall below the grid resolution, their properties are averaged out or completely ignored. However, these small features are far from insignificant; they can dominate fluid flow pathways, control sweep efficiency, and determine the success or failure of Enhanced Oil Recovery (EOR) projects. Sub-cell heterogeneity is the “hidden architecture” of the reservoir. When you fail to represent it, your reservoir model becomes a mere cartoon—a simplified, often misleading, representation of the complex reality of the subsurface.

Figure 1: The Scale Mismatch

The Deadly Mistakes of Ignoring Small-Scale Heterogeneity

The decision to ignore small-scale heterogeneity is not a neutral act; it actively introduces systematic errors into reservoir models. The consequences are far-reaching and can be financially catastrophic.

1. Over-Optimistic Recovery Estimates: Ignoring thin shale laminations in fluvial or deltaic reservoirs can grossly overestimate vertical connectivity. Simulators may predict efficient sweep, but in reality, fluid flow can be compartmentalized by centimeter-scale barriers. The result is an overestimation of recoverable reserves and expensive surprises when actual production falls short.
2. Incorrect Well Placement: Without accurate heterogeneity modeling, horizontal wells drilled in reservoirs dominated by inclined heterolithic stratification (IHS) may encounter unexpected water breakthrough. This can lead to misplaced wells, early coning, and bypassed hydrocarbons, significantly reducing the well’s profitability.
3. EOR Failures: Enhanced Oil Recovery processes, such as waterflooding, steam flooding, or polymer injection, rely on accurate predictions of sweep efficiency. Neglecting cross-bedding permeability anisotropy can be the difference between a successful project and outright failure. For instance, a 2008 study by Ringrose and Nordahl highlighted the importance of identifying the Representative Elementary Volume (REV) for permeability in heterolithic deposits, demonstrating how crucial accurate fine-scale modeling is for reliable predictions of fluid flow [cite: 1.1].
4. Poor Upscaling: The process of upscaling—moving from a fine-grid geological model to a coarser dynamic model—is essential in reservoir simulation. If sub-cell heterogeneity is ignored during this process, upscaling loses the very flow features that dominate reservoir dynamics. This leads to biased transmissibility multipliers and a distorted history match, which in turn compromises the predictive power of the model.

In essence, ignoring small-scale heterogeneity doesn’t just introduce noise—it introduces a systematic and often costly error that can jeopardize the entire project.

Figure 2: Homogeneous vs. Heterogeneous Fluid Flow

SBED: The Missing Link Between Core and Reservoir Simulation

Origins: SBED was developed in the early 2000s by Geomodeling Technology Corp. (now SubsurfaceAI Inc.) in collaboration with major oil companies. The motivation was clear: traditional geocellular models could not represent bedding-scale heterogeneities, yet core and outcrop studies consistently showed their overwhelming control on fluid flow. SBED filled this critical gap by offering a practical, physics-based workflow.

Concept: SBED models the geometry of sedimentary bedding structures—such as cross-bedding, lamination, or ripple sets—and calculates their flow-effective properties. It transforms a geologic description into upscaled petrophysical tensors that honor sub-grid variability.

Unique Contribution: SBED’s value proposition is built on several key innovations:

• It bridges the scale gap between plug-scale measurements and reservoir simulation scale.
• It allows geoscientists to directly translate sedimentological interpretation into reservoir flow properties.
• It seamlessly integrates into mainstream static modeling platforms like Petrel and RMS.
• It remains one of the few tools specifically designed to operationalize sedimentology in reservoir models.

The tool was designed not as a standalone scientific experiment but for practical, real-world application, ensuring that the complex task of modeling heterogeneity is made accessible to reservoir teams.

The SBED Workflow for Modeling Sub-Cell Heterogeneity

SBED’s success lies in its ability to simplify a highly complex problem. A typical workflow involves the following steps:

1. Sedimentological Input: The process begins with a geological description, which can come from core photos, core logs, or outcrop analogues. Key facies and structures are identified, such as laminated sands and shales, cross-bedded sets, inclined heterolithic stratification (IHS), and ripple lamination. SBED provides a comprehensive library of bedding templates that can be parameterized for thickness, proportion, and geometry to match the specific reservoir characteristics.
2. Rock Property Assignment: Petrophysical properties (porosity, permeability, saturation) are assigned to each lithofacies. These properties are typically derived from core plugs, well logs, or analogue studies.
3. Stochastic Realization: SBED uses a stochastic process to generate multiple realizations of the bedding-scale architecture. This is crucial for ensuring that the full range of geological variability is captured, rather than relying on a single deterministic model.
4. Flow Simulation at Sub-Cell Scale: This is a key innovation of the SBED process. The software conducts fine-scale flow simulations within the modeled bedsets. This step allows SBED to calculate effective permeability tensors and transmissibility multipliers that accurately represent the directional anisotropy of fluid flow induced by the bedding structures. A study by Theting et al. (2005) on pore-to-field multiphase upscaling for a depressurization process highlighted the importance of this kind of detailed analysis, demonstrating how SBED helps bridge this scale gap effectively [cite: 1.1].
5. Upscaling: The effective properties are then upscaled to the grid cell scale. Instead of assigning a single, isotropic permeability value, the model now includes anisotropic tensors that genuinely honor the reservoir’s heterogeneity. As demonstrated by Ringrose et al. (2004) in their work on permeability rescaling for the Heidrun field, this approach is vital for accurate near-wellbore modeling [cite: 1.1].
6. Export to Full-Field Models: The final results are seamlessly exported to mainstream reservoir modeling platforms. SBED does not disrupt existing workflows; it enhances them by providing high-fidelity data that improves the accuracy of the overall model.

Figure 3: An evolving SBED Workflow, from research to the mainstream applications. (A) Classic SBED workflow developed in the SBED JIP from 2000-2010. (B) A ML powered SBED workflow available in subsurfaceAI 2025

Integration with Full-Field Reservoir Models

SBED was intentionally designed for practical integration with existing industry workflows.

• Static Model Integration: The outputs from SBED—including effective properties and rock types—feed directly into geomodels built in platforms such as Petrel and RMS.
• Dynamic Simulation: Upscaled tensors and transmissibility multipliers generated by SBED can be directly incorporated into flow simulators like Eclipse and CMG.
• History Matching: By more accurately representing the reservoir’s underlying geology, incorporating SBED results often simplifies the history matching process by reducing the need for arbitrary multipliers.
• Forecasting: A more accurate representation of sweep efficiency leads to robust and reliable production forecasts, which are essential for making sound business decisions.

This tight integration ensures that modeling sub-cell heterogeneity is not merely an academic exercise but a process that delivers tangible, measurable value to reservoir teams.

Case Studies: When SBED Made the Difference

The impact of SBED is best illustrated through its real-world applications, as documented in various industry publications.

1. North Sea Tidal Reservoir: In a case study involving a North Sea tidal reservoir, ignoring thin shale drapes led to an overestimation of recovery factors by 15%. Incorporating SBED corrected the vertical transmissibility, leading to a much tighter history match and more accurate production forecasts. This aligns with the findings of Martinius et al. (2005), who discussed the reservoir challenges posed by heterolithic tidal sandstone reservoirs in mid-Norway and the importance of detailed characterization [cite: 1.1].
2. Athabasca Oil Sands: For Steam-Assisted Gravity Drainage (SAGD) projects in the Athabasca Oil Sands, performance was grossly overestimated when cross-bedding anisotropy was ignored. SBED modeling revealed preferential steam pathways, leading to improved SAGD performance forecasts. This demonstrates how SBED helps in complex, unconventional reservoirs where small-scale structures have a dominant impact on recovery processes.
3. Fluvial Reservoirs: In meandering river deposits with IHS, neglecting heterogeneity resulted in incorrect predictions of early water breakthrough. SBED provided anisotropy multipliers that corrected the sweep efficiency, ultimately saving millions in misplaced infill wells. This confirms the findings of Alpak and van der Vlugt (2014), who examined the dynamic impact and flow-based upscaling of estuarine point-bar stratigraphic architecture, showing how fine-scale structures dictate fluid movement and well performance [cite: 1.1].
4. Tidal Deltaic Reservoirs: The work of Nordahl et al. (2005) on a heterolithic tidal reservoir interval, where they used a process-based modeling tool, and Ringrose et al. (2005) on vertical permeability in the same setting, both published in Petroleum Geoscience, further demonstrate the effectiveness of SBED [cite: 1.1]. These studies highlight how a detailed, geologically informed approach to modeling bedding-scale features is essential for accurate petrophysical characterization and understanding of flow dynamics.
5. Thin-Layered Reservoirs: Cozzi et al. (2007) and Scaglioni et al. (2006) published work on processing core data and developing petrophysical characterization for thin-layered reservoirs, emphasizing the role of tools like SBED [cite: 1.1]. Their work underscores the industry’s long-standing challenge in accurately modeling these complex systems and the necessity of specialized software to overcome it.

With the release of subsurfaceAI 2025, the SBED workflow has included a machine-learning enhanced workflow (Figure 3b), upon the classic research oriented approach developed in the SBED JIP from year 2000-2010 (Figure 3a).

The Consequences of Not Modeling Sub-Cell heterogeneity

Failing to model sub-cell heterogeneity is not a neutral choice; it actively biases models toward optimism. The practical consequences are severe:

• Reserves: You risk booking barrels that do not exist, leading to inflated asset valuations.
• Production Strategy: You may overcommit to expensive facilities or pipelines based on overly optimistic production forecasts.
• Investor Trust: Inaccurate forecasts can severely damage the credibility of a company in the eyes of investors and stakeholders.
• Recovery Efficiency: Wells may be drilled that never pay back the initial investment, a multi-million or billion-dollar mistake.

These are not just academic errors; they are fundamental flaws in the decision-making process that can lead to significant financial losses. As a real world example, Hern (2011) reported more accurate and geologically consistent history matching after applying SBED derived vertical permeability in the static model for a sand reservoir. There are as much as 25% improvement in the production rate forecasting (Figure 4).

Figure 4: SBED derived vertical permeability can significantly improve the accuracy of history match for a matured sand reservoir, there are as much as 25% of improvement in the oil forecasting

SBED and the Future of SubsurfaceAI

Today, SubsurfaceAI continues to support SBED while expanding its portfolio with cutting-edge AI-driven workflows. SBED remains a cornerstone of its offerings, as it addresses a problem that no amount of seismic data or machine learning can bypass: the physical reality of bedding-scale heterogeneity.

As AI-driven reservoir modeling evolves, SBED’s approach—which honors geology at the finest resolvable scale—remains fundamental. SubsurfaceAI’s broader mission of integrating data and automating workflows ensures that SBED will continue to complement new technologies such as seismic AI and core analysis AI, creating a more holistic and accurate picture of the subsurface. The synergy between traditional geological modeling and modern AI techniques is the key to unlocking the full potential of reservoir characterization.

Conclusion

Reservoir models are only as robust as the geology they represent. Ignoring sub-cell heterogeneity is not a minor oversight but a systematic error that distorts forecasts, reserve estimates, and field development decisions. SBED provided the industry with the first practical, sedimentologically grounded solution to bridge this critical scale gap. By translating bedding-scale geological interpretations into simulation-ready properties, SBED ensures that reservoirs are modeled realistically and responsibly. The ultimate lesson is clear: if you don’t model sub-cell heterogeneity, you don’t truly know your reservoir. The wealth of published research and successful case studies serves as a powerful testament to the indispensable role of this modeling approach in modern petroleum engineering.