MACHINE LEARNING MODELS FOR ACCURATE MAGNITUDE ESTIMATION IN EARTHQUAKE EARLY WARNING SYSTEMS

Abstract

Earthquake early warning (EEW) systems play a vital role in minimizing disaster impacts by providing alerts within seconds after seismic events. Accurate and rapid magnitude estimation is a critical component of these systems. Traditional seismological techniques often suffer from latency and reduced accuracy for largemagnitude earthquakes due to waveform saturation. To address these limitations, this study explores the application of machine learning (ML) models for real-time magnitude estimation using early seismic waveform features. A dataset of seismic signals was preprocessed to extract time– frequency domain features, including peak ground acceleration, predominant period, spectral ratios, and waveform energy. Various machine learning models, including Random Forest, Support Vector Regression (SVR), and deep learning architectures such as Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, were trained and validated. Performance was evaluated using mean absolute error (MAE), root mean square error (RMSE), and correlation coefficient (R²). Results demonstrated that deep learning models, particularly CNNLSTM hybrids, achieved superior accuracy in magnitude estimation compared to conventional regression-based approaches, with reduced prediction latency. The findings highlight the potential of ML-driven frameworks in improving EEW system reliability, enabling faster decisionmaking and better disaster preparedness