📄 earth_science

Development of Earthquake Early Warning (EEW) System: Empirical relationship using data from Japan

This study utilizes strong motion data from 143 Japanese earthquakes (2010–2024) to establish empirical relationships between various P-wave parameters and earthquake magnitude, demonstrating that a 5-second analysis window combined with multiple parameters significantly enhances the reliability of magnitude estimation for Earthquake Early Warning systems.

Original authors: R Gowrinanda, Utpal Saikia, Atul Saini, Himanshu Mittal

Published 2026-07-06

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: R Gowrinanda, Utpal Saikia, Atul Saini, Himanshu Mittal

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 Picture: The Seismic "Heads-Up" System

Imagine you are standing in a room when a heavy bookshelf starts to fall. You hear a small creak (the P-wave) before the shelf actually crashes down (the S-wave). If you could instantly analyze that creak to know exactly how heavy the shelf is and how hard it will hit, you could jump out of the way before the crash.

This is exactly what an Earthquake Early Warning (EEW) system tries to do. It listens to the first, fast-moving "creak" of an earthquake (the P-wave) to predict the size of the disaster before the slow, destructive shaking arrives.

The Problem: Guessing the Size Too Soon

Scientists have been trying to figure out how big an earthquake is just by listening to those first few seconds of the P-wave. However, it's like trying to guess the volume of a symphony orchestra by listening to only the first three notes of a violin. Sometimes you get it right; sometimes you think a small tune is a massive symphony, or vice versa.

The researchers in this paper wanted to find the best "mathematical recipe" to make that guess as accurate as possible. They didn't just look at one clue; they looked at a whole toolbox of different measurements.

The Ingredients: A Toolbox of Clues

The team used data from 143 earthquakes in Japan (a place that is very good at recording earthquakes). They looked at 10 different "ingredients" or parameters to see which ones gave the best clues about the earthquake's size (Magnitude).

Think of these ingredients like different ways to measure a storm:

Pa, Pv, Pd (Peak Acceleration, Velocity, Displacement): These measure the instant "kick" of the wave. Imagine measuring how hard a single raindrop hits a window.
PGA, PGV, PGD: These are the maximum "shakes" recorded. Like measuring the highest wave in a storm.
τc (Characteristic Time): This measures how long the "creak" lasts. A short creak might mean a small event; a long creak might mean a big one.
CAA (Cumulative Absolute Absement) & CAV (Cumulative Absolute Velocity): These are the "total effort" measurements. Instead of just looking at the hardest hit, they add up every bit of movement and shaking over time. It's like measuring the total weight of all the rain that fell, not just the biggest drop.
RSSCV: This is a special "cumulative score" that adds up the speed of the shaking in all directions.

The Experiment: The 3-Second vs. 5-Second Rule

The researchers tested these ingredients over three different time windows: 3 seconds, 4 seconds, and 5 seconds after the earthquake starts.

The 3-second window: This is like trying to guess the size of a movie after watching only the trailer. It's fast, but the data is messy and scattered.
The 5-second window: This is like watching the first five minutes of the movie. You have more information, so the pattern becomes much clearer.

The Finding: The 5-second window was the winner. It provided the clearest picture with the least amount of "static" or confusion.

The Results: Who Won the Race?

When they compared all the ingredients to see which one predicted the earthquake's size most accurately, here is what they found:

The Old Favorite (Pd): Traditionally, scientists have relied heavily on Pd (Peak Displacement). It's like the "classic" tool everyone uses. It works okay, but it has a flaw: for very big earthquakes, it tends to "saturate." Imagine a thermometer that stops rising at 100°F even if the fire is actually 200°F. Pd stops giving accurate numbers for massive quakes.
The New Champion (CAA): The study found that CAA (Cumulative Absolute Absement) was the most reliable. Because it adds up all the movement over time, it doesn't get "confused" by huge earthquakes. It keeps growing linearly with the size of the quake, giving a more accurate prediction.
The Dark Horse (RSSCV): Another parameter, RSSCV, also performed incredibly well. Like CAA, it didn't get "saturated" by big earthquakes. It kept a steady, straight-line relationship with the earthquake's size, making it a very strong candidate for future warning systems.
The Noise-Sensitive One (τc): The "time duration" parameter (τc) was very sensitive to background noise. It was like trying to hear a whisper in a noisy room; the results were often unreliable.

The Conclusion: A Multi-Tool Approach

The paper concludes that while we can predict earthquake sizes quickly, we shouldn't rely on just one number.

The Best Time: Waiting 5 seconds after the first P-wave arrives gives the most accurate data.
The Best Tools: Instead of just using the old "Peak Displacement" (Pd), we should use a combination of tools, specifically CAA and RSSCV. These tools are better at handling massive earthquakes without getting "stuck" or underestimating the danger.

In short: By listening to the first 5 seconds of an earthquake and adding up all the movement (not just the biggest jolt), we can build a smarter, more reliable alarm system that tells us exactly how big the earthquake is before the bad shaking hits.

Technical Summary: Development of Earthquake Early Warning (EEW) System: Empirical relationship using data from Japan

Problem Statement
Earthquake Early Warning (EEW) systems are critical for mitigating the destructive impacts of seismic events by providing alerts before damaging S-waves arrive. While systems in Japan and Taiwan have demonstrated efficiency, a significant challenge remains in the rapid and accurate estimation of earthquake magnitude using only the initial phases of P-wave arrival. Current methodologies often rely on single or dual parameters (such as peak displacement, $P_d$ , or characteristic time, $\tau_c$ ), which can suffer from saturation at higher magnitudes ( $M_w > 7.0$ ) or high sensitivity to signal-to-noise ratios (SNR). This study addresses the need to evaluate a multi-parameter approach to enhance the robustness, reliability, and timeliness of magnitude estimation for EEW systems.

Methodology
The study utilizes a comprehensive dataset comprising 143 earthquake events recorded by the National Research Institute for Earth Science and Disaster Prevention (NIED) K-NET network in Japan between 2010 and 2024. The dataset includes moment magnitudes ( $M_w$ ) ranging from 5.0 to 7.7, with hypocentral distances up to 110 km and focal depths up to 60 km, totaling 4,446 waveforms.

The methodology involves extracting ground motion parameters from the initial P-wave arrivals using three-component waveforms. The parameters analyzed include:

Peak values: Peak acceleration ( $P_a$ ), velocity ( $P_v$ ), and displacement ( $P_d$ ).
Ground motion indices: Peak Ground Acceleration (PGA), Velocity (PGV), and Displacement (PGD).
Cumulative and frequency-based parameters: Cumulative Absolute Absement (CAA), Cumulative Absolute Velocity (CAV), Characteristic time period ( $\tau_c$ ), and Root Sum of Squares Cumulative Velocity (RSSCV).

These parameters were calculated for time windows of 3, 4, and 5 seconds following the P-wave arrival. A fourth-order Butterworth band-pass filter (0.075 Hz low-cut) was applied to remove noise and drift. Linear regression analyses were performed to establish empirical relationships between these parameters and both magnitude ( $M$ ) and hypocentral distance ( $R$ ), following the formulation: $\log(P) = a \cdot \log(R) + b \cdot M + c$ . The Standard Deviation Error (SDE) of the regression was calculated to determine the most reliable time window and parameters for magnitude prediction.

Key Results

Time Window Optimization: The analysis indicates that the 5-second time window consistently yields the lowest scattering and the most stable correlations for all parameters compared to 3-second and 4-second windows.
Parameter Correlations:
- Strong linear correlations (near 1:1) were observed between PGA and $P_a$ , PGV and $P_v$ , and PGD and $P_d$ specifically in the 5-second window.
- CAA and RSSCV Superiority: The study found that Cumulative Absolute Absement (CAA) and RSSCV exhibit the strongest relationships with magnitude. CAA demonstrated the lowest Standard Deviation Error (SDE) of 0.356 in the regression against magnitude for the 5-second window.
- Saturation Behavior: Traditional amplitude-based parameters ( $P_a, P_v, P_d$ ) and CAV showed signs of saturation for magnitudes greater than 7.3, leading to potential underestimation. In contrast, CAA and RSSCV maintained linear trends without significant saturation even at higher magnitudes.
- $\tau_c$ Limitations: While $\tau_c$ did not saturate, it exhibited high sensitivity to SNR and significant data scattering, resulting in predicted magnitudes with SDE values insufficient for reliable early warning.
Empirical Equations: The study derived specific regression equations for the 5-second interval relating $P_a, P_v, P_d, CAA, CAV, \tau_c,$ and $RSSCV$ to magnitude and distance. For instance, the relationship for CAA is given as: $\log_{10}(CAA_5) = -1.408 \log_{10}(R) + 0.602 M - 0.092$ .

Significance and Claims
The paper claims that a multi-parameter approach, specifically incorporating CAA and RSSCV alongside traditional metrics, offers a more robust foundation for EEW systems than single-parameter methods. The authors assert that:

CAA is a superior magnitude indicator: It provides a more stable and reliable estimate of earthquake magnitude (SDE of 0.48 for CAA vs. 0.53 for $P_d$ ) and captures the total ground motion field more comprehensively than vertical-component-only displacement.
RSSCV prevents underestimation: As a cumulative energy-dependent parameter, RSSCV avoids the saturation issues common in amplitude-based parameters, making it effective for estimating the size of large earthquakes ( $M > 6.0$ ).
5-Second Window is Optimal: The study concludes that a 5-second interval after P-wave arrival is the most suitable timeframe for extracting these parameters to balance prediction accuracy with warning lead time.

The authors emphasize that these findings lay the groundwork for real-time magnitude prediction using multiple earthquake parameters from the initial P-wave phases, potentially reducing false alarms and improving the accuracy of EEW alerts. They suggest that future work could further integrate these parameters and explore machine learning approaches to enhance system accuracy.