Phenomenological Diagnosis of Seismic Regimes.

This methodological report presents an open, model-independent protocol for the phenomenological diagnosis of seismic regimes based on the Phenomenological Reconstruction of Complex Systems (PRCS/FRSS) framework. Instead of seeking deterministic predictions of individual earthquakes, the method diagnoses phase regimes of a seismically active medium by reconstructing its internal dynamics directly from catalogue data. Using the Kamchatka earthquake catalogue (1990–2026, Ml ≥ 3.5, depth 0–70 km), we aggregate discrete events into a daily energy series s(t), reconstruct the phase space via Takens' embedding (m=4, τ=3 days), and compute dynamic invariants: correlation dimension D₂, flux balance J, phase risk metric Φ, and coherence C. Regime clustering in the invariant space identifies windows of elevated structural risk. A combined criterion (Φ ≥ 0.7 ∧ J ≥ 0.1) isolates a narrow class of pre-critical states with a true alarm rate of ~94% and false alarm rate of ~6% within a 90-day horizon for events Ml ≥ 6.0. Wavelet analysis of risk windows reveals a systematic morphological difference: successful windows exhibit a compact, post-centre power maximum on a narrow scale range, while false alarms show diffuse, pre-centre activity. This two-stage scheme (phase invariants + spectral verification) enables discrimination between trajectories leading to macro-events and those dissipating into background. The protocol is fully reproducible: all computations are implemented as open Python scripts. The method is designed as a supervisory diagnostic layer compatible with existing seismic monitoring systems and PSHA models. It does not replace operational early-warning systems but provides analysts with a tool for medium-term assessment of structural risk and regime evolution.