Density estimation of weak periodic signals in pre-earthquake seismic waves

Abstract

We introduce a method for identifying weak periodic components in pre-earthquake seismic waveforms by examining the scale-index response of a driven Duffing chaotic oscillator. This nonlinear setup helps detect and classify subtle deterministic features buried in low-amplitude, noisy seismic records. We apply this approach to seismic data collected before three moderate-to-strong earthquakes, and compare the results with two quiescent control periods. The weak periodic signals (WPS) detected by the approach exhibit clear, systematic frequency shifts. Kernel density estimates highlight these changes in the dynamics of the Marmara fault segment south of Istanbul, the likely location of a future large Istanbul earthquake. The results also suggest that chaos-based detection methods can reveal weak periodic patterns with potential applications in real-time seismic monitoring and earthquake forecasting. More generally, this study adds to the growing interaction between nonlinear dynamics and geophysics by providing another perspective on the complex behaviour of active fault systems. The density-based analysis of WPSs proposed here may also be useful in other fields that involve large, noisy datasets with chaotic characteristics, such as physiology or financial systems.