📄 earth_science

Topology-Phase Monitoring of Kilauea Eruption Preparation: A Reproducible Seismic Audit and Forecasting Potential Assessment

This study demonstrates that a reproducible topology-phase monitoring approach applied to Kilauea's seismic data successfully identified the 2018 eruption preparation 21 days in advance by detecting a distributed reorganization of seismicity, outperforming traditional scalar metrics like event count and energy.

Original authors: Guojun Pan

Published 2026-07-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Guojun Pan

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

The Big Idea: Listening to the "Shape" of Earthquakes

Imagine you are trying to predict when a crowded room is about to erupt into a chaotic mosh pit.

Usually, scientists try to predict volcanic eruptions by counting how many people are shouting (earthquake counts) or how loud the shouting is (energy). But the author of this paper, Guojun Pan, suggests a different approach. He argues that before a volcano erupts, the earthquakes don't just get louder or more frequent; they start to organize.

Think of it like a crowd of people.

Normal days: People are scattered randomly, chatting in small, disconnected groups.
Eruption preparation: People start moving together, forming a single large line, clustering tightly in one corner, and moving in the same direction.

This paper tests whether we can spot that "reorganization" in the Kilauea volcano's earthquake data before the 2018 eruption happened.

The Experiment: A "Seismic Audit"

The researcher treated the Kilauea volcano like a patient and performed a "reproducible audit." Here is how they did it:

The Data: They looked at public earthquake records from 2014 to 2019 within an 80km radius of Kilauea.
The Training: They studied the earthquakes from 2014–2017 to learn what "normal" looks like for that time of year.
The Test: They watched the data from early 2018 (leading up to the May 3rd eruption) to see if the pattern changed.

Instead of just counting earthquakes, they turned the data into a 7-day rolling movie. For every week, they built a "map" (a graph) connecting earthquakes that happened close to each other in space and time.

The "Topology" Score: Measuring the Shape

The paper introduces a new way of measuring the volcano's health called Topology-Phase Monitoring.

Imagine the earthquakes are dots on a piece of paper.

Normal State: The dots are scattered like rain on a window. They don't touch much.
Eruption State: The dots start connecting with strings, forming a giant, dense web that moves in one direction.

The researcher created a "Topological Score" that measures:

Connectivity: Are the dots linking up into one big group?
Density: Is the web getting tighter?
Anisotropy: Is the web stretching in a specific direction (like a line) rather than being a round blob?
Migration: Is the center of the web moving?

The Results: The Early Warning System

The study found something very interesting when they compared this new "Shape Score" against the old "Count Score":

The Old Way (Counting): If you just counted how many earthquakes happened, the alarm didn't go off until 2 days before the eruption.
The New Way (Topology): The "Shape Score" crossed its danger threshold 21 days before the eruption.

The Analogy:
Imagine a storm approaching.

The Count method is like waiting until you see the first heavy raindrop on your roof to say, "It's raining!" (Too late).
The Topology method is like noticing the clouds have stopped drifting randomly and have formed a tight, dark wall moving toward you. You know a storm is coming long before the first drop falls.

What This Means (and What It Doesn't)

What the paper claims:

The 2018 Kilauea eruption was preceded by a distinct "reorganization" of earthquakes. The seismic activity didn't just get louder; it got more structured and connected.
This structural change was detectable three weeks earlier than simple earthquake counts.
This suggests that monitoring the shape of seismic activity is a better way to get an early warning than just counting quakes.

What the paper does NOT claim:

It is not a crystal ball. The author explicitly states this method cannot predict the exact date, time, or size of an eruption. It only signals that the volcano is entering a "preparation state."
It is not a magic bullet for all volcanoes. This test was done on Kilauea (a very well-monitored volcano) using only earthquake data. The author admits that for a real-world warning system, we need to add other data like ground deformation (GPS), gas levels, and heat.
It is not a guarantee. The method detected the state of preparation, but it cannot tell you if the volcano will definitely erupt or if the magma will just stop moving underground.

The Bottom Line

This paper proposes that volcanoes "get their act together" before they blow. By watching how earthquake patterns connect and organize (rather than just how many there are), scientists might be able to get a much earlier warning—potentially weeks instead of days. However, this is currently a research finding, not a finished product, and it needs to be tested on other volcanoes and combined with other sensors before it can be used for real-time public safety.

Technical Summary: Topology-Phase Monitoring of Kilauea Eruption Preparation

Problem Statement
High-precision volcanic eruption forecasting remains a significant challenge because eruption preparation is rarely signaled by a single scalar variable (e.g., a spike in earthquake count). Instead, it involves a distributed reorganization of seismicity, deformation, gas transport, and magma-fracture connectivity. Traditional monitoring often relies on isolated channels, which can be ambiguous; for instance, earthquake swarms do not always lead to eruptions, and increased event counts may reflect stress adjustment rather than imminent activity. The paper posits that eruption preparation may be better understood as a transition in the "topology-state" of the seismic field—shifting from diffuse background activity to a connected, migrating, and anisotropic state.

Methodology
The study performs a reproducible, seismic-only audit of the Kilauea volcano system using public USGS FDSN earthquake catalogue data.

Data Scope: The analysis covers an 80 km radius around Kilauea (19.4069°N, 155.2834°W) for earthquakes with magnitude $M \ge 1.0$ .
Temporal Windows: Data was aggregated for same-calendar windows (January 1 to May 5) from 2014 to 2019. The 2014–2017 windows served as the training reference distribution, while the 2018 window was the target test case for the lower East Rift Zone (LERZ) eruption (onset: May 3, 2018).
Topology-State Vector: A seven-day rolling window was converted into a vector of seven features:
1. Event count ( $N$ ).
2. Magnitude-energy proxy ( $E$ ).
3. Largest connected-component fraction ( $C_{max}$ ).
4. Graph edge density ( $\rho_G$ ).
5. Normalized spatial entropy ( $S_{sp}$ ).
6. Event-cloud anisotropy ( $A$ ).
7. Centroid migration ( $\Delta r_c$ ).
- Graph Construction: Events were connected if separated by $\le 8$ km in projected space and $\le 2$ days in time.
Indicator Construction: A composite Topology-Phase Indicator (TPS) was calculated as the mean of standardized components. Spatial entropy was negated to reflect the transition from diffuse to concentrated organization. Thresholds were set at the 99th percentile of the 2014–2017 training distribution.

Key Results
The audit, using locked parameters trained on 2014–2017 data, yielded the following findings regarding the 2018 LERZ eruption:

Early Warning: The composite topology-phase indicator first crossed the training 99th-percentile threshold on April 12, 2018, providing a 21-day lead time before the May 3 eruption onset.
Comparison with Scalar Metrics:
- Event Count: Crossed its 99th percentile only on May 1, 2018 (2 days before onset).
- Energy Proxy: Did not cross its threshold before the onset.
Component Analysis: The early warning was driven by a coordinated structural reorganization. Graph density breached the threshold on April 12, while the largest connected-component fraction and spatial entropy breached their respective thresholds on April 17. Migration showed an early breach in March but did not exceed the training maximum until May 1. Anisotropy breached early in January but was deemed insufficient as a standalone predictor.
Peak Values: The topology indicator reached a peak value of 4.0411 on May 1, 2018, exceeding both the training 99th percentile (2.0462) and the training maximum (2.8166).

Key Contributions

Topology-State Detection: The study demonstrates that eruption preparation in a well-monitored basaltic system can be detected as a topological transition in the seismic field, distinct from simple increases in event count or energy.
Reproducible Audit: The authors provide a fully reproducible workflow, including specific scripts and data files, allowing for the independent verification of the "locked-parameter" audit.
Lead Time Improvement: The results suggest that combining connectivity, density, and spatial organization metrics can significantly extend the lead time for detecting eruption-preparation states compared to traditional scalar baselines.

Significance and Claims
The paper makes a modest, specific claim: it does not assert that topology-state physics can deterministically predict the exact date, vent location, or magnitude of all volcanic eruptions.

Probabilistic Forewarning: The primary contribution is the validation of a probabilistic, multi-channel detection of an "eruption-preparation state." The method identifies when a system has entered an unusually organized topology state relative to historical baselines.
Operational Target: The appropriate near-term target is not deterministic prediction but the improvement of lead time and the reduction of scalar false alarms through topology-state monitoring.
Limitations and Future Steps: The authors acknowledge that the current audit is seismic-only and relies on a single strong case (Kilauea 2018). They state that for the method to advance to a high-precision operational claim, it requires:
- Validation on independent non-eruptive unrest intervals to estimate false alarm rates.
- Integration of multi-channel data (deformation, gas, thermal, tilt).
- Comparison against standard models (e.g., b-value, ETAS).
- Strict parameter locking before testing on new events.

In conclusion, the paper supports the hypothesis that Kilauea's 2018 eruption preparation manifested as a detectable topological reorganization of seismicity, offering a potential pathway for earlier, probabilistic state detection in volcanic monitoring.