📄 earth_science

An Upward-Continuation-Based Framework for Subsurface Structural Reconstruction from Gravity Anomalies

This study introduces the UCRD–NFG framework, a model-independent method that combines upward-continuation-based decomposition with the Normalized Full Gradient technique to efficiently reconstruct subsurface structural geometry and estimate depths directly from Bouguer gravity anomalies without requiring inversion or prior geological information.

Original authors: Ako Alipour, Khalil Motaghi, Zahra Mousavi

Published 2026-07-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ako Alipour, Khalil Motaghi, Zahra Mousavi

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Earth's crust is like a giant, layered cake, but instead of seeing the layers, you can only feel the "weight" of the cake from the very top. In geophysics, this "weight" is measured as gravity anomalies. The problem is that this weight is a messy mix of everything underneath: shallow rocks, deep mountains, hidden caves, and dense ore bodies all pulling on the scale at once. It's like trying to figure out exactly what's inside a wrapped gift just by lifting it; you know it's heavy, but you can't tell if it's a book, a brick, or a bowling ball without opening it.

This paper introduces a new way to "unwrap" that gift without actually digging a hole or doing complex, computer-heavy math. The authors call their method UCRD (Upward Continuation Restriction to Decomposition), and they pair it with a depth-finder called NFG.

Here is how it works, broken down into simple steps:

1. The "Foggy Camera" Trick (Upward Continuation)

Imagine you are taking a photo of a landscape through a thick fog. The closer you are to the ground, the more details you see (trees, rocks), but the background is blurry. If you float up in a hot air balloon, the small details disappear, and you start to see the big shapes of the mountains in the distance.

The UCRD method does exactly this, but with math instead of balloons. It takes the gravity data and mathematically "floats" the observation point higher and higher into the sky.

Low altitude: You see the "short waves" (shallow, small objects).
High altitude: The short waves fade away, and you are left with the "long waves" (deep, massive objects).

By doing this at many different heights, they can separate the gravity signal into different "layers" based on how deep the source is.

2. The "Enhanced Filter" (Analytic Signal & Normalization)

Once they have separated the layers, the images are still a bit fuzzy and hard to read. Think of this step like using a photo editor to sharpen the edges and adjust the contrast.

They use a mathematical tool called the Analytic Signal to highlight the edges of the objects (like tracing the outline of a shape).
Then, they apply a Scale-Adaptive Normalization. This is like adjusting the brightness and color of each layer so they all look consistent. Crucially, this step keeps the "polarity" (whether the object is heavy or light compared to its surroundings) and gives the image real-world units (like milligals), so scientists can actually measure it.

The result is a set of clear, colored maps called SRWL (Scaled Restricted Wavelength). These maps show the shape and outline of the underground structures, layer by layer, from shallow to deep.

3. The "Depth Gauge" (NFG)

While the UCRD maps tell you what the shape looks like, they don't tell you exactly how deep it is (because the "balloon" height is just a math trick, not a real depth).

To fix this, they use a second tool called NFG (Normalized Full Gradient). Think of NFG as a specialized depth gauge or a sonar. It analyzes the data to find the exact depth where the "signal" is strongest. It doesn't draw the shape, but it tells you, "The object you see in the UCRD map is sitting at 10 kilometers deep."

4. Putting It All Together

The paper tested this "UCRD + NFG" team-up on three fake underground models:

Isolated Cubes: Like scattered Lego blocks at different depths.
Overlapping Blocks: Like two heavy suitcases stacked on top of each other, making a confusing lump.
A Giant Stepped Structure: Like a massive, wide staircase stretching for miles.

The Results:

The UCRD successfully drew the outlines of all these shapes, even when they were overlapping or very deep. It could separate the "Lego blocks" from the "suitcases" and show the "staircase" clearly.
The NFG accurately told them how deep each part was.
Together, they created a 3D picture of the underground world without needing to know the geology beforehand or running slow, expensive computer simulations.

Why This Matters (According to the Paper)

Usually, figuring out what's underground requires either:

Guessing: Making a model based on what you think the rocks look like (Forward Modeling).
Heavy Lifting: Running massive computer programs that try millions of possibilities to find the best fit (Inversion).

This new framework is different. It's like having a quick, model-free sketch. You don't need to know the geology in advance, and you don't need a supercomputer. It gives you a clear, physical picture of the underground shapes and their depths immediately. The authors suggest this is perfect for getting a "first look" at an area before deciding to do more expensive, detailed studies later.

In short: They built a mathematical "X-ray" that separates the Earth's gravity signal into depth layers, sharpens the edges to show shapes, and uses a depth gauge to tell you how deep those shapes are—all without needing to dig or guess.

Technical Summary: An Upward-Continuation-Based Framework for Subsurface Structural Reconstruction from Gravity Anomalies

Problem Statement
Gravity data interpretation is fundamental for investigating subsurface density heterogeneity but is inherently non-unique; substantially different density distributions can generate similar anomaly responses. While inversion techniques and forward modeling exist, they often rely heavily on prior geological constraints, complex parameterization, or computationally intensive optimization. Transformation-based methods (e.g., upward continuation, analytic signal, normalized full gradient) have been developed to extract structural information without inversion, yet most focus on anomaly enhancement, boundary detection, or wavelength separation rather than reconstructing the full spatial geometry of causative bodies while preserving anomaly polarity. Furthermore, gradient-based methods like the Normalized Full Gradient (NFG) are effective for edge detection and depth estimation but do not directly represent the complete geometry or density polarity of the source. There is a need for a model-independent framework that can reconstruct subsurface structural geometry directly from gravity anomalies, providing both geometric characterization and depth constraints without requiring external geological data.

Methodology
The authors propose the Upward Continuation Restriction to Decomposition (UCRD) framework, integrated with the Normalized Full Gradient (NFG) method, to address these limitations. The workflow consists of three primary stages:

Multi-Level Upward Continuation and Spectral Decomposition:
The Bouguer gravity anomaly field is decomposed into restricted wavelength (RWL) levels using multi-level upward continuation. The process begins with the observed field ( $U_0$ ) and applies successive upward continuations ( $U_n$ ) with exponentially increasing heights ( $\Delta z_n$ ) to isolate progressively deeper sources. The restricted wavelength level for each step is calculated as the difference between the field at the previous height and the current height ( $RWL_n = U_{n-1} - U_n$ ). This separates the gravity field into components associated with specific depth intervals.
Analytic Signal Transformation and Scale Normalization:
To enhance structural continuity and preserve physical interpretability, each RWL component undergoes a three-dimensional analytic signal transformation. A scale-adaptive normalization procedure is then applied to restore physical units (mGal) and preserve anomaly polarity. This involves calculating a level-specific scale factor based on the dynamic range of the RWL and the analytic signal amplitude. The result is a Scaled Restricted Wavelength (SRWL) representation. These SRWL maps capture the magnitude and direction of lateral density contrasts at each depth level, providing a pseudo-3D representation of the anomaly evolution.
Integration with NFG for Depth Constraints:
Since upward continuation does not yield direct physical depth information, the UCRD geometric reconstruction is complemented by the NFG method. The NFG is applied using a Fourier-series formulation with signal-mirror extension to avoid edge discontinuities. An adaptive procedure determines the optimal number of harmonics ( $N$ ) by tracing maximum NFG responses across increasing harmonic intervals. This provides quantitative depth constraints that are integrated with the UCRD structural geometry.

Key Contributions

UCRD Framework: Introduction of a decomposition method that transforms gravity anomalies into physically interpretable images of subsurface structural geometry without inversion.
SRWL Representation: Development of a scaling procedure that restores physical units and anomaly polarity to analytic signal results, enhancing the visibility of structural features obscured in raw data.
Complementary Workflow: A unified approach where UCRD delineates structural geometry and polarity, while NFG supplies independent depth constraints, reducing the ambiguity inherent in gradient-based interpretations.
Model Independence: The methodology operates directly on gravity data, requiring no prior geological information, seismic data, or computationally intensive optimization.

Results
The framework was validated using three synthetic models of increasing complexity:

Isolated and Stepped Bodies: The method successfully delineated six cubic bodies and three stepped structures with varying depths and density contrasts. UCRD reconstructed the spatial distribution and polarity, while NFG identified corresponding depths. The results showed strong agreement with the true model, even for deep bodies partially masked by shallow anomalies.
Overlapping Sources: In a model with partially overlapping prisms, UCRD successfully separated the dominant structural components of shallow and deep sources, resolving the composite anomaly signatures. NFG provided stable depth estimates for both the shallow and deep components, distinguishing overlapping geometries that are difficult to interpret via standard gravity analysis.
Laterally Extensive Structures: For a model comprising twenty-five blocks distributed at various depths, UCRD recovered the overall geometry of the large-scale structure, revealing both shallow and deep components. NFG identified distinct depth levels and structural boundaries within the extensive source system.

In all cases, the SRWL transformations produced sharper and more localized anomaly signatures compared to standard restricted wavelength components. The combined UCRD–NFG workflow demonstrated the ability to recover both horizontal and vertical structural patterns across a broad range of source configurations.

Significance and Claims
The paper claims that the proposed UCRD–NFG framework offers a practical, computationally efficient tool for gravity interpretation. Its primary significance lies in its ability to provide a preliminary characterization of subsurface geometry and depth without the need for inversion or external constraints. The authors position the method not as a replacement for forward modeling or joint inversion, but as a valuable preliminary step that can generate initial models, geometric constraints, and parameterization guidance for subsequent, more intensive investigations. The framework is particularly highlighted for regional-scale investigations where rapid structural characterization is required. The study acknowledges limitations inherent to spectral decomposition, such as partial wavelength overlap between levels and smoothing effects from vertical interpolation, but asserts that the method provides a robust and intuitive approach for structural imaging.