Subsurface fault geometries in inland Hokkaido from geometric clustering of dense hypocenter distributions

Beneath the surface of the Earth, the ground is not a solid, unbroken shell. Instead, it is fractured by massive cracks called faults, where tectonic plates grind against one another. When stress builds up along these cracks and finally releases, the result is an earthquake. For scientists trying to predict where and how strongly the ground might shake, knowing the exact shape and direction of these hidden cracks is essential. However, while geologists can map faults that break the surface, the vast majority of dangerous earthquakes occur on "blind faults" that never reach the top. These invisible structures lie deep underground, making them nearly impossible to see with traditional tools like satellite imagery or surface surveys. Without a clear picture of these hidden geometries, it is difficult to understand how stress moves through the crust or to accurately assess the risk of future disasters.

In the northern Japanese island of Hokkaido, this challenge is particularly acute. The region sits at a complex junction where several massive tectonic plates collide and slide past each other. This tectonic activity generates frequent earthquakes, yet many of the faults responsible for them remain unknown because they are buried deep beneath the soil or hidden by thick layers of snow and vegetation. To solve this puzzle, a team of researchers turned to the earthquakes themselves. By analyzing the precise locations of thousands of small tremors that have occurred over the last two decades, they were able to reconstruct the three-dimensional shapes of the faults that caused them. Their work provides the first comprehensive map of these hidden structures across the entire inland region of Hokkaido, revealing a landscape of deep, steep cracks that differ significantly from what surface maps had suggested.

The researchers began by listening to the Earth with an incredibly sensitive ear. They gathered continuous recordings from a dense network of seismometers scattered across Hokkaido, capturing every vibration from 2004 to 2024. Using advanced computer models trained to recognize the faint signals of earthquakes, they identified hundreds of thousands of seismic events that traditional methods had missed. Many of these were tiny tremors, too small to be felt by humans, but together they formed a detailed cloud of points marking where the ground had broken. The team then refined the location of each event with extreme precision, correcting for the fact that seismic waves travel at different speeds through different types of rock. This process resulted in a highly accurate map of where the earthquakes happened, effectively turning the seismicity into a glowing outline of the invisible fault lines.

With this dense cloud of earthquake locations in hand, the team applied a new method to find the flat planes that represent the faults. They looked for patterns in the way the points clustered together, treating the collection of earthquake locations like a cloud of dust that naturally settles into the shape of the cracks they slid along. By analyzing the orientation of these clusters, they could determine the direction and steepness of the faults. The result was a reconstruction of 184 distinct rectangular fault segments across inland Hokkaido. These findings revealed that the region is dominated by steep, high-angle faults, most of which lie between 6 and 12 kilometers deep. This depth is significant because it places the earthquakes firmly within the upper crust, the brittle layer of the Earth's interior where rocks break rather than flow.

The map they created tells a clear story about the tectonic forces shaping the island. In central Hokkaido, the faults form long, north-south lines that align perfectly with the boundary where two major geological regions are colliding. This collision zone, known as the Hidaka collision zone, is where the Kuril arc is pushing against the Northeast Japan arc. The researchers found that the faults in this area are steep and well-organized, suggesting that the crust is being squeezed and broken in a very specific, predictable way. In eastern Hokkaido, the pattern shifts. Here, the faults trend northeast to southwest, mirroring the movement of a sliver of crust that is being pushed sideways by the subducting Pacific plate. These orientations match the broader motion of the tectonic plates, confirming that the hidden faults are responding directly to the forces driving the region's geology.

One of the most striking discoveries was the depth of the faults in different areas. While most of the seismic activity occurred in the upper crust, the researchers found deep faults extending below 20 kilometers, but only in the southern part of central Hokkaido. This area corresponds to the thickest part of the crust, where the collision between the tectonic arcs has piled up the rock. In contrast, the volcanic regions in the north and west showed very shallow faulting, likely because the heat from magma and hot fluids weakens the rock, preventing deep cracks from forming. The study also highlighted areas where the fault patterns are complex and difficult to define, such as the northernmost tip of the island, where slow, creeping movements of the crust may be creating a chaotic mix of fractures.

To test the reliability of their new map, the researchers compared their reconstructed faults against the known rupture patterns of eight major earthquakes that had occurred in the region over the last two decades. They found that for six of these events, the faults they identified matched the actual rupture planes with remarkable accuracy, differing by only a few degrees. This included the devastating 2018 earthquake in eastern Iburi, where their model correctly identified the steep, east-dipping fault that caused the disaster. This agreement gives scientists confidence that their method can successfully reveal the geometry of faults even before a large earthquake strikes. It also suggests that the hidden faults they mapped are the same ones that will likely produce future earthquakes.

The study also looked at how likely these faults are to slip again under the current stress conditions. By calculating the forces acting on each fault, the researchers found that the steep faults along the central collision boundary are perfectly aligned to fail, meaning they are under high stress and ready to break. In contrast, the faults running along the volcanic chain in the northeast and southwest showed a much lower tendency to slip. This suggests that in these volcanic areas, local conditions—such as high pressure from underground fluids or variations in the stress field—are preventing the faults from failing, or perhaps changing the way they behave. This distinction is crucial for hazard assessment, as it helps identify which parts of the region are most vulnerable to immediate rupture.

Despite the success of the new map, the researchers acknowledge its limits. The method relies entirely on the presence of dense seismic activity; in areas where few earthquakes occur, such as the Ishikari lowland or the Tokachi plain, the faults remain invisible. This means that while the map covers a vast portion of the island, it does not yet show every fault system. Additionally, the assumption that faults are flat, rectangular planes is a simplification; in reality, faults are often curved and complex. However, the researchers argue that this approach provides a solid first-order estimate, offering a level of detail that was previously impossible to achieve. By turning the invisible tremors of the Earth into a clear geometric picture, this work offers a new foundation for understanding the seismic risks in Hokkaido and provides a model for how similar hidden structures might be mapped in other tectonically active regions around the world.