🔬 physics

Physics-Based ETAS Modeling of Persistent Reservoir-Induced Seismicity at Song Tranh 2, Vietnam

This study employs an integrated physics-based ETAS framework to analyze 14 years of reservoir-induced seismicity at Vietnam's Song Tranh 2, revealing that while natural tectonic background dominates, persistent activity is driven by fault interactions and rapid fluid transport through anisotropic basement faults that lower frictional thresholds to reactivate stable structures.

Original authors: Trong Cao Dinh, Tuan Thai Anh, Trieu Cao Dinh, Bach Mai Xuan

Published 2026-07-31

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Trong Cao Dinh, Tuan Thai Anh, Trieu Cao Dinh, Bach Mai Xuan

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 not as a solid, unyielding rock, but as a giant, ancient sponge sitting deep underground. Sometimes, this sponge is squeezed by the massive weight of tectonic plates shifting against each other, building up invisible pressure like a coiled spring waiting to snap. This is the world of seismology, the study of earthquakes. Usually, these springs snap on their own when the pressure gets too high, causing a natural earthquake. But what happens when we pour a giant bucket of water onto that sponge? This is the puzzle of "Reservoir-Triggered Seismicity." When humans build huge dams and fill them with water, the added weight and the water seeping into cracks can sometimes act like a trigger, causing the Earth to shake sooner than it would have otherwise. Scientists have been trying to figure out exactly how much of the shaking is caused by the water itself versus how much is just the Earth's natural脾气 (temper) coming out. Understanding this is crucial because it helps us keep dams safe and protects people living nearby from unexpected tremors.

Now, picture the Song Tranh 2 hydroelectric dam in Vietnam. Since the reservoir was filled in 2010, it has been shaking more than 8,000 times over 14 years. That's a lot of rattling! A team of researchers from the Vietnam Academy of Science and Technology decided to build a super-smart digital model to figure out exactly why this was happening. They used a method called "Physics-based ETAS," which is a fancy way of saying they created a computer simulation that breaks down every single earthquake into four different "ingredients" to see which one is the main chef.

Think of the earthquake activity like a pot of soup. The researchers wanted to know: Is the soup spicy because of the chili (the water pressure)? Is it salty because of the ocean (the natural tectonic stress)? Is it hot because of the stove (the weight of the water)? Or is it bubbling because the pot is shaking itself (the earthquakes triggering other earthquakes)?

Here is what their digital taste-test revealed:

First, they found that the biggest ingredient, making up about 63.59% of the shaking, is the Natural Tectonic Background. This means the Earth in this area was already full of built-up stress, like a spring that was already wound up tight. The reservoir didn't create new energy; it just acted like a little nudge that let the spring release its energy a bit early. The dam didn't cause the earthquake; it just helped it happen sooner.

Second, they discovered that Fault Interaction (or earthquakes triggering other earthquakes) accounts for 27.10% of the activity. This is the "domino effect." Once the first few quakes happened, they shifted the stress around, causing more quakes to follow, even when the water level stayed perfectly still. This explains why the shaking has persisted for so long, long after the initial filling of the dam.

Third, the water itself played a role, but in two different ways. The Pore Pressure (water seeping deep into the cracks) contributed 8.07%. This is the delayed trigger. The water didn't just sit on top; it soaked into the deep rock, lubricating the faults and making them slippery. The researchers found something amazing here: the rock beneath the dam isn't the same in all directions. It's like a stack of wooden planks where water flows easily down through the cracks but struggles to move sideways. They calculated that water can travel 11.29 times faster vertically (straight down) than horizontally. Because of this "super-conductor" effect, the water managed to reach depths of 5.5 km in just 48 days, triggering quakes much faster than scientists used to think was possible.

Finally, the Elastic Loading (the immediate weight of the water pressing down) was the smallest ingredient, contributing only 1.24%. This means the simple act of the water getting heavy wasn't the main reason for the shaking; the slow seepage of water was far more important.

The team also solved a mystery: some of the faults in the area looked "locked" and stable, like a door that shouldn't open. But their model showed that the deep water pressure reduced the friction on these doors, allowing them to slide open even though they seemed too strong to move.

In short, this study didn't just guess; they ran a complex mathematical simulation using 8,273 recorded earthquakes to prove that the Song Tranh 2 earthquakes are a mix of the Earth's natural stress being released early, a chain reaction of quakes triggering more quakes, and water acting as a fast-traveling lubricant deep underground. Their model is so reliable that even when they tweaked the numbers slightly, the results stayed the same, giving them high confidence in their findings. This helps us understand that while the dam didn't create the energy, it definitely helped the Earth let it out, and the way water moves through the rocks is much faster and more directional than we previously imagined.

Technical Summary: Physics-Based ETAS Modeling of Persistent Reservoir-Induced Seismicity at Song Tranh 2, Vietnam

Problem Statement
The Song Tranh 2 (ST2) hydropower reservoir in Vietnam represents a globally significant case of persistent Reservoir-Triggered Seismicity (RTS). Since impoundment in 2010, the region has experienced over 8,000 seismic events, including magnitudes up to $M_L$ 4.7, with activity continuing for over 14 years even during periods of stable water levels. While previous studies have identified fluid diffusion and elastic loading as potential triggers, and noted the presence of NW-SE trending fault systems, there remains a lack of quantitative differentiation regarding the proportional contributions of distinct physical mechanisms. Specifically, the relative roles of instantaneous elastic response, delayed pore pressure diffusion, fault-to-fault stress interactions, and the natural tectonic background in sustaining this prolonged seismic sequence require rigorous decoupling.

Methodology
The study employs an Integrated Physics-based ETAS (Epidemic Type Aftershock Sequence) framework to analyze a catalog of 8,273 seismic events recorded between 2012 and 2020. The workflow integrates five core steps:

3D Geomechanical Framework: Seismic events and fault geometries were standardized to the UTM Zone 49N coordinate system. The actual 3D distance from the reservoir bed to each hypocenter was calculated to determine stress components.
Hydraulic Anisotropy Quantification: Using the Upper Envelope Fitting method, the study decoupled hydraulic diffusivity coefficients ( $D$ ) for distinct fault systems, distinguishing between sensitive NW-SE corridors and locked W-E systems.
Physics-Based ETAS Decomposition: The total seismicity rate $\lambda(t)$ $λ (t)$ was modeled as a linear combination of four independent physical components:
- Elastic Loading ( $E(t)$ ): Instantaneous Coulomb stress changes ( $\Delta CFS$ ) derived from 3D Boussinesq integration of reservoir water level variations.
- Pore Pressure Diffusion ( $P(t)$ ): Delayed triggering simulated via a 3D Green's function for a semi-infinite medium, incorporating site-specific anisotropic diffusion coefficients ( $D_x, D_z$ ).
- Fault Interaction ( $O(t)$ ): Self-triggering stress transfer modeled using the Modified Omori Law with a decay exponent $p=1.05$ .
- Tectonic Background ( $\bar{N}$ ): A constant stochastic rate representing natural endogenous seismicity of the Kon Tum block.
Optimization and Inversion: A Sequential Least Squares Programming (SLSQP) algorithm minimized the loss function between observed and predicted seismic rates to determine the optimal physical weights ( $w_i$ ) for each component, subject to constraints ( $0 \le w_i \le 1, \sum w_i = 1$ ).
Validation: Model robustness was assessed through sensitivity analysis across varying friction coefficients ( $\mu = 0.60–0.75$ ) and focal depths (4.5–6.5 km), alongside Cross-Correlation Function (CCF) analysis.

Key Results

Quantitative Mechanism Decoupling: At an anchor depth of 5.5 km, the inversion results reveal that natural tectonic background activity is the dominant driver, accounting for 63.59% of the seismicity. This confirms the reservoir acts as an external trigger releasing pre-accumulated endogenous energy rather than generating new energy.
Persistence Mechanism: Fault-to-fault stress interaction contributes 27.10%, identified as the primary factor driving the sequence's multi-year persistence through cascading self-triggering, even when water levels are stable.
Triggering Contributions: Pore pressure diffusion accounts for 8.07%, acting as the external hydraulic trigger that destabilizes deep faults. Elastic loading contributes minimally at 1.24%, indicating the instantaneous response is not the dominant mechanism.
Hydraulic Anisotropy: The study validates strong basement hydraulic anisotropy. The vertical diffusivity ( $D_z$ ) is 11.29 times higher than the horizontal component ( $D_x$ ). This high vertical conductivity allows fluid pressure to penetrate to hypocentral depths (5.5 km) within 48 days.
Stress Field and Fault Reactivation: An optimized regional stress field with a maximum principal compressive stress axis ( $\sigma_1$ ) of 320° explains the clustering of epicenters along the NW-SE corridor. Slip tendency analysis shows primary faults (e.g., F1, F8) have a low initial slip tendency ( $T_s \approx 0.36$ ), classifying them as "locked" under dry-rock conditions. The study demonstrates that fluid pressure reduces the effective normal stress, lowering the frictional threshold to promote failure despite this low initial stability.

Significance and Claims
The paper claims to provide a robust quantitative platform for seismic hazard assessment by successfully decoupling the complex interactions governing persistent RTS. By integrating 3D geomechanics with stochastic modeling, the framework resolves the mechanism of fluid-driven reactivation along stable faults and quantifies the specific role of hydraulic anisotropy in accelerating fluid transport. The authors assert that the model's stability under varying input parameters (with weight variations less than 5%) confirms its reliability. Consequently, this approach offers a vital tool for risk assessment and dam safety management for hydropower projects with similar geological conditions globally, moving beyond empirical observations to a physics-based understanding of seismogenic structures.

More like this