Universal scaling between precursory duration and event size across mechanically driven geohazards

By applying a physics-based framework to a global dataset of 109 geohazard events, this study reveals a universal, near-linear scaling relationship between precursory duration and failure volume, indicating that catastrophic failures across diverse mechanically driven systems are governed by common organizing mechanisms involving the progressive growth of correlated deformation toward a dynamical critical point.

The Gist

Here is an explanation of the paper using simple language and creative analogies.

The Big Idea: Finding the "Warning Clock" for Disasters

Imagine you are watching a balloon being slowly inflated. At some point, you know it's going to pop. But the big question is: When did the "pop" process actually start? Was it when the balloon was the size of a grape? A grapefruit? Or just seconds before it burst?

For decades, scientists have struggled to answer this for real-world disasters like landslides, rockfalls, glacier collapses, and volcanic eruptions. They knew these events usually speed up (accelerate) right before they happen, but they couldn't agree on when that speeding-up phase began. Was it a week ago? A year ago? The answer depended on who you asked and how they measured it.

This paper introduces a new, objective "physics-based stopwatch" that tells us exactly how long the warning period lasts, and it reveals a surprising secret about how nature works.

---

1. The Problem: Guessing the Start Time

Previously, scientists looked at data (like how fast a rock is moving or how much gas a volcano is spitting out) and tried to guess when the "acceleration" started. They used rules of thumb, like "if the speed doubles, that's the start."

The Analogy: Imagine trying to find the start of a song by listening to a radio station with static. You might say, "I think the music started when the volume got loud," but that's subjective. One person might say it started at the first drumbeat; another might say it started when the singer began. Because everyone used different "rules," they couldn't compare different disasters to see if they followed the same pattern.

2. The Solution: The "Smart Fit" (LPPLS)

The authors used a sophisticated mathematical model called LPPLS (Log-Periodic Power Law Singularity). Think of this model not as a ruler, but as a smart detective.

Instead of guessing when the acceleration started, the model tries to fit the data to a specific curve that describes how things break. It asks: "If I assume the acceleration started on Day 1, does the math work? What if I assume it started on Day 10? Day 50?"

It uses a special "penalty system" (called Lagrange regularization) to prevent cheating. If the model tries to say the acceleration started too early (when the data was actually just random noise), the penalty is high. If it starts too late, the fit is bad. The model finds the perfect balance—the one specific day where the math makes the most sense without forcing it.

The Result: This gives them an objective "Start Time" for the warning phase for 109 different disasters across the globe.

3. The Big Discovery: The "Size-Speed" Rule

Once they had the start times for 109 events (from tiny rockfalls to massive landslides and volcanoes), they looked for a pattern. They compared the size of the disaster (how much rock or ice fell) with the length of the warning time.

The Analogy: Imagine you are running a race.

• If you are running a 100-meter dash, you might start sprinting 10 meters before the finish line.
• If you are running a marathon, you might start sprinting 10 kilometers before the finish line.

The paper found that for mechanically driven disasters (landslides, rockbursts, glacier breaks), there is a strict rule: The bigger the disaster, the longer the warning.

Specifically, they found a "Universal Scaling Law." If you double the size of the rockslide, the warning time doesn't just double; it increases in a predictable, mathematical way. It's like nature has a built-in timer that is directly linked to the size of the object about to break.

• Small rockfall (a few tons): Warning might be a few days.
• Huge landslide (millions of tons): Warning might be a few months.
• Massive glacier collapse: Warning might be a year or more.

4. The "Volcano" Exception

Interestingly, this rule did not work for volcanoes.

The Analogy: Think of a landslide like a stack of dominoes. Once the first domino falls, the whole chain reaction is inevitable and follows a set path based on how many dominoes there are. The bigger the stack, the longer the chain reaction takes to reach the end.

Volcanoes, however, are more like a kettle boiling on a stove. The water (magma) might boil over quickly if the heat is turned up high, or slowly if the heat is low. The size of the kettle (the volcano) doesn't strictly dictate how long it takes to boil; it depends on the heat source (magma pressure) and other complex factors. Because of this, volcanoes didn't follow the same "size = time" rule as landslides and rockfalls.

5. Why This Matters: The "System-Wide" View

The most profound insight is why this rule exists.

For a long time, people thought a landslide starts at a tiny crack deep underground, and we just can't see it until it gets big. This paper suggests that's wrong.

The Analogy: Imagine a crowd of people in a stadium.

• Old View: A riot starts with one angry person in the back, and the panic spreads slowly to the front.
• New View (from this paper): The whole stadium is getting restless at the same time. The "warning phase" is the time it takes for that restlessness to spread and synchronize across the entire crowd until they all move together.

The paper argues that for landslides and rockfalls, the "warning phase" is the time it takes for the entire unstable mass to organize itself. The bigger the mass, the longer it takes for the "signal" to travel from one side to the other and synchronize the collapse.

Summary

• The Tool: They built a math model that objectively finds the start of a disaster's warning phase, removing human guesswork.
• The Rule: For landslides, rockfalls, and glaciers, bigger disasters have longer warning times in a predictable way.
• The Lesson: These disasters aren't just a small crack getting bigger; they are the entire system "warming up" and organizing before the final snap.
• The Benefit: This helps scientists predict how much lead time they might have. If they know the size of a potential landslide, they can estimate how long the warning phase will last, helping communities prepare better.

In short, nature has a rhythm, and this paper finally figured out the beat.

Technical Summary

Here is a detailed technical summary of the paper "Universal scaling between precursory duration and event size across mechanically driven geohazards" by Qinghua Lei and Didier Sornette.

1. Problem Statement

Forecasting catastrophic geohazards (e.g., landslides, rockbursts, glacier breakoffs, volcanic eruptions) remains a significant challenge. While many such events are preceded by an observable phase of accelerating deformation or seismicity, the duration of this precursory phase is poorly constrained.

• Current Limitations: Previous studies typically define the onset of acceleration using heuristic thresholds, visual inspection, or empirical criteria tied to specific observables. This introduces subjectivity, making it difficult to compare precursory durations across different events, scales, or hazard types.
• Core Question: Is there a universal relationship between the duration of the precursory phase and the size of the eventual failure? If so, what physical mechanisms govern this relationship?

2. Methodology

The authors propose a physics-based, objective framework to quantify precursory duration, moving away from subjective thresholding.

• The LPPLS Model: The study utilizes the Log-Periodic Power Law Singularity (LPPLS) model. This model describes the approach to a critical point (failure) by capturing both the overall power-law acceleration and superimposed log-periodic oscillations (reflecting discrete scale invariance and hierarchical damage accumulation).
  • Formula: $\Omega(t) = A + B(t_c - t)^m + C(t_c - t)^m \cos[\omega \ln(t_c - t) + \phi]$
  • Where $\Omega$ is the observable (displacement, energy, etc.), $t_c$ is the critical time, $m$ is the critical exponent, and $\omega$ is the log-periodic frequency.
• Objective Onset Identification (Lagrange Regularization):
  • Instead of fixing the start time of the precursory phase, the authors treat the start time ($\tau$) as an endogenous variable.
  • They employ a Lagrange regularization approach to minimize a cost function: $\chi^2(\tau, \theta) = r^2(\tau, \theta) + \lambda \cdot p(\tau)$.
  • Here, $r^2$ is the sum of squared residuals, $p$ is the number of parameters, and $\lambda$ is a Lagrange multiplier (analogous to a chemical potential in statistical physics) that penalizes excessively short calibration windows.
  • This method objectively identifies the onset time ($\tau^*$) that balances model fit with window length, eliminating the need for a priori assumptions.
• Dataset: The framework was applied to a global dataset of 109 geohazard events spanning seven continents over the past century. The dataset includes:
  • 49 Landslides (rockslides, soilslides, rockfalls)
  • 11 Rockbursts
  • 17 Glacier breakoffs (including ice shelves)
  • 32 Volcanic eruptions
  • Note: Earthquakes were excluded due to ongoing debates regarding the universality of accelerating seismic release.

3. Key Contributions

1. Objective Quantification: The study introduces a rigorous, physics-based method to define precursory duration without relying on arbitrary thresholds, allowing for consistent comparison across diverse geohazards.
2. Universal Scaling Discovery: The authors identify a robust scaling law between precursory duration ($T$) and failure volume ($V$) specifically for mechanically driven instabilities (landslides, rockbursts, glaciers).
   • The relationship follows a power law: $T \propto V^{\zeta}$, with an exponent $\zeta \approx 0.35 \pm 0.05$.
   • When expressed in terms of characteristic system size ($L \propto V^{1/3}$), the relationship becomes nearly linear: $T \propto L^{1.05}$.
3. Differentiation of Mechanisms: The study explicitly demonstrates that volcanic eruptions do not follow this scaling (exponent $\approx 0$), highlighting a fundamental physical distinction between mechanically driven failures and magmatic processes.

4. Results

• Scaling Analysis:
  • For mechanically driven events (excluding volcanoes), the data collapses onto a clear trend spanning more than ten orders of magnitude in failure volume.
  • The scaling exponent $\zeta = 0.35$ implies that the "effective preparatory rate" ($\gamma^* = L/T$) is approximately 0.85 m/day (with a wide confidence interval reflecting system variability).
  • Statistical tests (OLS, Likelihood Ratio, Permutation, Bootstrap) decisively reject the null hypothesis of no scaling for mechanically driven hazards.
• Volcanic Anomaly: Volcanic eruptions show no significant correlation between duration and volume. Their precursory durations fall within the confidence bounds of the mechanical scaling but do not follow the trend, suggesting different controlling mechanisms (e.g., magma transport, phase changes, rheology) rather than progressive system-spanning damage accumulation.
• Case Studies: The method was validated on four distinct case studies (Veslemannen landslide, NSW rockburst, Grandes Jorasses glacier, Axial Seamount volcano), successfully identifying onset times that align with physical observations and converging to stable plateaus as more data becomes available.

5. Significance and Implications

• Finite-Size Criticality: The near-linear scaling between system size ($L$) and precursory duration ($T$) supports a finite-size scaling interpretation. It suggests that catastrophic failure is a critical phenomenon where correlated deformation grows until it spans the entire system. The precursory duration is determined by the time required for these correlations to reach the system scale, rather than by local rupture kinetics.
• Monitoring Strategy: The findings challenge the view that surface measurements are merely proxies for subsurface processes. Instead, they suggest that surface monitoring directly samples the system-scale precursory dynamics in mechanically driven instabilities. This validates the empirical success of surface-based forecasting (e.g., radar, GNSS) for landslides and glaciers.
• Predictive Framework: By establishing a universal scaling law, the study provides a theoretical basis for estimating the "actionable lead time" for early warning systems based on the estimated size of the unstable mass.
• Mechanistic Insight: The divergence between mechanically driven hazards and volcanic eruptions underscores that while the mathematical description of acceleration (LPPLS) may be similar, the physical drivers (damage accumulation vs. magmatic pressurization) differ fundamentally.

In conclusion, this paper provides a unifying physical framework that links the temporal organization of failure to the spatial scale of the hazard, offering a robust tool for understanding and potentially forecasting catastrophic geohazards.