A Petrel-Based Reproducible Workflow for Integrated Static Reservoir Characterization Using Synthetic Seismic and Well-Log Data

Deep beneath the earth's surface, oil and gas companies rely on three-dimensional maps to understand where resources are hidden. These maps are not photographs taken with a camera; they are complex hypotheses built from scattered clues. One set of clues comes from seismic data, which uses sound waves to create a broad, blurry picture of rock layers over a wide area. Another set comes from well logs, which are detailed measurements taken from narrow holes drilled into the ground, offering high-resolution information but only along a single vertical line. To make sense of the space between these sparse measurements, geologists use geological concepts to fill in the gaps. The result is a static reservoir model, a digital representation of the underground rock that helps engineers estimate how much oil or gas exists, plan where to drill next, and predict how fluids will move. However, building these models involves countless human decisions, from how to interpret a wavy line on a screen to how to smooth out rough data. When these choices are not written down clearly, the final map becomes difficult to check or repeat, leaving the science vulnerable to hidden errors and making it hard for others to learn from the work.

In a recent study, researcher Innocent J. Aluka from Prairie View A&M University addresses this problem by introducing a new way to organize the creation of these underground maps. The study does not invent a new way to calculate rock properties or discover a new type of oil reservoir. Instead, it proposes a structured, step-by-step framework called the Reproducible Integrated Reservoir Modeling Workflow, or RIRMW. This framework treats the creation of a reservoir model not as a loose collection of software clicks, but as a strict chain of auditable steps. Each step in the chain requires specific inputs, follows a clear transformation rule, records every parameter used, passes a quality check, and produces a versioned output before moving to the next stage. The goal is to make the entire process transparent, ensuring that every judgment call and setting is documented so that another expert could follow the same path and arrive at the same result.

To test this idea without using confidential data from real oil fields, the researcher built a complete, synthetic dataset. This artificial environment mimics a faulted, layered rock formation containing sandstone, shaly sandstone, and shale. The team generated fake seismic images, attribute maps that highlight rock boundaries, and detailed well logs that simulate what a drill would encounter. Using this controlled data, they ran the new workflow through six distinct modules. First, they interpreted the seismic and well data to identify rock layers and faults. Second, they built a structural grid, a digital skeleton that holds the rock layers in place while respecting the faults that cut through them. Third, they simulated the distribution of rock types, deciding where sandstone and shale appear in the 3D space. Fourth, they modeled continuous properties like porosity, which measures how much empty space exists in the rock to hold fluid, and permeability, which measures how easily fluid can flow through it. Fifth, they integrated all these pieces into a single, consistent 3D model. Finally, they assessed uncertainty by running the process multiple times with different random seeds to see how much the results might vary.

The results of this demonstration showed that the workflow successfully produced a coherent, internally consistent model of the synthetic reservoir. The final map showed sandstone bodies connected in a way that matched the geological concepts, with high porosity and permeability concentrated in the better-quality rock zones. Crucially, the process preserved a complete record of every decision made. For example, the study recorded that the effective porosity ranged from zero to forty percent, with an average of 17.6 percent across the different simulated scenarios. The horizontal permeability, a measure of flow, ranged from 0.1 to 1,000 millidarcies, a unit used to describe how easily fluids move through rock. Water saturation, which indicates how much of the pore space is filled with water rather than oil, was modeled between zero and one. By running multiple simulations, the team could show that while the exact arrangement of rock varied slightly each time, the overall statistics remained stable, with a median porosity of 17.3 percent.

A key finding of the study is that this approach does not eliminate the need for expert judgment; rather, it makes that judgment visible. In traditional projects, a geologist might make a choice about how to connect two data points, but that choice often remains a mental note or a quick sketch. In this new framework, that choice is recorded as a specific parameter in a register. If the modeler decides to use a particular mathematical rule to estimate permeability based on porosity, that rule is written down along with the specific numbers used. This allows anyone reviewing the work to see exactly how the model was built and to understand why it looks the way it does. The study explicitly notes that because the software used is proprietary, it is impossible to guarantee that two different computers will produce the exact same number for every single cell in the grid. However, the framework ensures that the process is repeatable and that the logic behind the results is clear.

The research also highlights the importance of uncertainty. Because the data from the ground is sparse, there is no single "correct" map of the underground. Many different arrangements of rock could fit the same data. The workflow addresses this by generating an ensemble of multiple possible models, or realizations, rather than just one. This allows scientists to see the range of possibilities, from the most optimistic scenario to the most pessimistic one. In the synthetic test, the researchers showed that while the specific location of a sandstone channel might shift between different runs, the overall volume and connectivity of the rock remained within a predictable range. This helps decision-makers understand the risks involved, knowing that the model represents a spectrum of possibilities rather than a single, potentially misleading truth.

Ultimately, this work serves as a bridge between the complex, often opaque world of industrial reservoir modeling and the principles of open, reproducible science. It demonstrates that even when using commercial software where the inner workings are hidden, the process of building a model can be made transparent. By formalizing the steps and recording every decision, the framework allows the scientific community to audit, learn from, and build upon the work of others. The study concludes that while the specific numbers generated in this test are based on artificial data and cannot be used to predict actual oil reserves, the method itself is robust. It provides a practical way to ensure that the complex digital models used to explore the Earth's resources are built on a foundation of clear logic and documented evidence, making the science behind the maps as reliable as the data that feeds them.