By Oriona Egwo

At the recently concluded 2024 American Geophysical Union (AGU) Fall Meeting held at Washington, D.C., Ifeanyi, a researcher at Penn State University presented his research findings to a global audience of experts. 

This scholarly presentation comes at a time when global energy systems are rapidly evolving—from mature oil reservoirs to geothermal wells and carbon storage sites. Yet, the industry continues to face a stubborn truth: we can only manage what we can see. In the subsurface system, millions of years of geological processes have shaped porous rock systems kilometers underground, creating intricate networks of flow pathways that control how fluids and heat move. 

For engineers and researchers like Ifeanyi, “seeing” means reconstructing these hidden pathways from sparse measurements and limited imagery. Ifeanyi Samson Nwankwo, a doctoral researcher at Penn State University, is tackling that challenge by training artificial intelligence to turn narrow microscopic snapshots into panoramic subsurface insights. The research he brought to the public glare at AGU 2024 Conference sits at the intersection of geoscience, data science, and energy engineering, especially on using AI-driven image reconstruction to accelerate how scientists characterize rocks and make field-scale decisions.

From Sparse Images to Deep Understanding

Modern X-ray micro-computed tomography (micro-CT) scanners can reveal the intricate pore architecture of rocks at micrometer resolution—but only for small samples that that cannot form a fair representation of porous media system. Larger cores, which are better representative samples and more relevant for reservoir-scale modeling, must be scanned at lower resolution, losing vital pore-scale details that impact flow behavior. Ifeanyi’s dissertation bridges this gap using generative deep learning models trained to perform super-resolution—the process of reconstructing fine-scale structures from coarse imagery. By learning the statistical and physical relationships between pore- and core-scale images, his models can infer realistic high-resolution representations of larger rock volumes. This innovation could drastically cut the time, cost, and uncertainty involved in extracting key rock properties—such as porosity, permeability, and connectivity—that underpin reservoir modeling, CO₂ sequestration design, and geothermal resource evaluation. Early outcomes from his research were featured at the 2024 American Geophysical Union (AGU) Fall Meeting, the world’s largest Earth and space science conference, signaling growing academic and industrial interest in data-driven subsurface modeling.

Why It Matters for the Energy Industry

Reliable petrophysical characterization is the foundation of all subsurface projects—from oil recovery and carbon storage to hydrogen and geothermal operations. Yet, scaling measurements from pore-sized rock plugs to entire reservoirs has long been a source of uncertainty. If AI can infer high-resolution structure from lower-resolution data, engineers can evaluate larger sample sets with greater precision—improving statistical confidence and reducing dependence on expensive imaging. The result is a faster, more scalable workflow for estimating flow-relevant parameters, enabling smarter site screening, more accurate reservoir forecasts, and better risk management in complex geological formations. This vision aligns with Ifeanyi’s broader mission: to make digital rock physics faster, cheaper, and more physically trustworthy—so that energy transition projects such as CO₂ sequestration and geothermal heat extraction can progress with greater confidence.

By enabling higher-fidelity rock characterization and scaling that capability to reservoir volumes, Ifeanyi’s research could unlock billions of dollars in savings and added value for U.S. subsurface energy projects. For example, the geothermal industry alone is forecast to be worth USD 9.81 billion in 2024, growing to USD 13.56 billion by 2030. Improvements in reducing uncertainty in CO₂ storage or geothermal forecasts by even a few percent would translate into hundreds of millions in avoided costs or increased returns across U.S. energy portfolios.

Pairing Artificial Intelligence with Domain Physics

While his research focuses on image super-resolution, the broader project of Ifeanyi’s work emphasizes “science-informed” machine learning—where data-driven models are guided by the physics of the system. At Penn State’s Department of Energy and Mineral Engineering, Ifeanyi has contributed to a U.S. Department of Energy (DOE) SMART Initiative project on science-informed AI for real-time subsurface decisions. Building on this foundation, his models embed physical constraints and geological priors into the learning process, reducing the risk of unrealistic predictions often seen in purely data-driven approaches. This makes the AI-generated images not only visually plausible but also physically consistent with flow dynamics and transport laws—a crucial step for integrating them into reservoir simulators or permeability upscaling workflows. “AI can fill in the details, but physics tells it what’s real,” Ifeanyi explains. “The goal isn’t just beautiful images—it’s trustworthy predictions that engineers can act on.”

Building Trust in AI for the Subsurface

Trust remains the defining barrier to AI adoption in critical infrastructure. In his scholarly presentation, Ifeanyi addresses this challenge through a three-pronged strategy:

1. Developing a custom super-resolution GAN architecture tailored to porous media analysis.

2. Designing evaluation metrics grounded in physical properties—porosity, coordination number, tortuosity—rather than purely visual similarity.

3. Linking outputs to flow properties, ensuring predictions honor the governing equations of fluid transport.

This combination of technical rigor and domain awareness mirrors a larger industry shift toward explainable and verifiable AI, especially in safety-critical applications like energy production and carbon management.

Well-Grounded Background

Before beginning his PhD, Ifeanyi earned a First-Class B.Tech. in Petroleum Engineering from Rivers State University, followed by an M.Sc. at the African University of Science and Technology (AUST), Abuja, where he graduated top of his class. His academic excellence paved the way to Penn State, where he now combines his passion for research with a commitment to teaching and mentorship.

As a Graduate Teaching Assistant, he has supported petroleum engineering courses and guided undergraduate projects, translating complex data-science methods into concepts students can grasp and apply. His earlier experiences supervising student research in Nigeria have shaped his philosophy of mentorship—one that balances technical depth with accessibility, encouraging young engineers to see AI not as a black box, but as a tool for creative problem-solving in real-world energy systems.

Ifeanyi’s work exemplifies the next frontier of subsurface science—one where AI and physics coexist, producing insights that are both data-rich and scientifically grounded. As energy transitions toward cleaner and more efficient systems, the ability to “see” beneath the surface with confidence will define the pace of innovation. “Every reservoir is a puzzle,” he reflects. “If AI can help us see the missing pieces, then we can make better choices—not just for energy, but for the planet.”