Earthquake Estimation Using Observation Images and ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional earthquake estimation methods struggle to promptly and accurately estimate the moment magnitude (Mw) of earthquakes, especially when multiple earthquakes occur simultaneously, leading to false warnings, as seen in the 2018 Japan Meteorological Agency's Earthquake Early Warning system.
Innovation Solution
An earthquake estimation method using a computer-based approach that generates observation images of seismic wave propagation from multiple observation points and employs machine learning models to estimate earthquake parameters, including the number and magnitude of earthquakes, through numerical simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If surface waves are used to estimate moment magnitude (Mw) of an earthquake, then the physical estimation of earthquake magnitude is improved, but the response time is delayed making it difficult to promptly estimate the Mw
Solution Approach 1:
The system performs preliminary analysis by generating observation images from P-wave and S-wave data immediately after earthquake occurrence, before surface waves arrive. The earthquake estimation model is pre-trained with simulated data to enable rapid magnitude estimation from body waves alone, eliminating the waiting time for surface waves while maintaining accuracy through the trained model's ability to infer Mw from earlier arriving body wave patterns.
2Device complexity
If conventional methods assume seismic waves are generated from one earthquake, then the estimation process is simplified, but the system fails to accurately handle simultaneous earthquakes leading to false warnings
Solution Approach 1:
The system segments the seismic wavefield into multiple independent earthquake sources by generating separate observation images for different hypocenter positions. The earthquake estimation model is trained to recognize and separate simultaneous earthquake signals by analyzing spatial distribution patterns in the observation images, allowing it to identify multiple distinct earthquake events rather than treating them as a single composite event.
Solution Approach 2:
The system transforms the problem from temporal analysis to spatial analysis by creating observation images that show spatial distribution of seismic wave propagation. This dimensional transformation allows the model to simultaneously analyze multiple earthquake sources in the spatial domain, distinguishing between simultaneous earthquakes based on their different spatial patterns rather than trying to separate them in the time domain.
3Measurement precision
If the number of observation points is increased to improve estimation accuracy, then the measurement precision is improved, but the data processing complexity and computational load increase
Solution Approach 1:
The system merges data from multiple observation points into a single comprehensive observation image that shows the spatial distribution of seismic wave propagation across the entire network. This consolidation allows the earthquake estimation model to process all observation point data simultaneously through a unified spatial analysis approach, maintaining the benefits of multiple sensors while avoiding the complexity of processing each point separately.
Data Source
AI summary
An earthquake estimation method for more promptly estimating an earthquake on the basis of observation data. The earthquake estimation method includes, by a computer: generating an observation image showing a spatial distribution of seismic wave propagation on a basis of an observation result of seismic waves at a plurality of observation points on a ground; and estimating a parameter of an earthquake with respect to the observation image by using an earthquake estimation model in which a parameter of an earthquake including at least a position of a hypocenter and a magnitude is associated with a simulated observation image showing a spatial distribution of seismic wave propagation on a ground obtained from a result of a numerical simulation of the earthquake, performed with the parameter.