Presentation Information
[S25-13]Real-Time Monitoring of Strain-Energy Accumulation Using Physical Wavelets: The 2025 Kamchatka M8.8 and Tokara Swarm
*Fumihide TAKEDA1,2 (1. Takeda Engineering Consultant, 2. Earthquake Prediction Institute)
The Kamchatka Peninsula lies along a major subduction zone where the Pacific Plate subducts beneath the Okhotsk (North American) Plate. These plates are coupled by megathrust faults that rupture in great earthquakes. Southward, the boundary extends into the Japan Trench and links with the Nankai and Ryukyu troughs. Both continental and oceanic lithosphere are represented by three thermo-mechanically coupled layers: a brittle upper crust (B), a ductile lower crust (D), and the ductile uppermost mantle. Plate-driving forces produce brittle deformation in the B layer and steady-state creep deformation in the D layers. The creep deformation is thermo-mechanically coupled with B-layer deformation across the lithosphere.
Deformation in the brittle crust arises from two sources: direct deformation in the B layer and creep-deformation transfer from the D layer (Zoback & Zoback, 2002). GNSS coseismic strain observations suggest that 80–90% of effective B-layer deformation is direct brittle deformation, while 10–20% reflects ductile creep-deformation transfer. This upward creep-deformation-induced stress accumulation across the D–B transition zone leads to slow, deterministic stress accumulation at the base of the B layer. Such creep-deformation-induced stress accumulation carries physically predictable information about the genesis process and strain-energy accumulation and release cycles of significant earthquakes across subduction zones.
For real-time monitoring, we analyzed seismic catalog data (NIED, 2025) across the Japan subduction zones (16°–52° N, 116°–156° E, 36° × 40° mesh) using Physical Wavelet (PW)-based methods. Seismic time-series data were constructed from events with magnitude >= Mc (=3–4), forming d(c,m) series where c represents latitude (LAT), longitude (LON), depth (DEP), inter-event time (INT), and magnitude (MAG). Here, m denotes the chronological order of events (time). Each series yields three independent active variables with 15–25% stochastic fluctuations, resembling Brownian motion; these variables represent the three principal stress components.
Each earthquake (EQ) is modeled as a virtual EQ particle of unit mass, with successive events tracing an evolving trajectory in stress-sensitive space. Displacement, velocity, and acceleration are extracted as Newtonian quantities by PW operators applied to otherwise stochastic time series. From cumulative INTs and DEPs over 30 consecutive EQs (typically 20–70) with M >= Mc (=4), we defined two normalized strain-energy density indicators: NCI(m,30) (Normalized Cumulative Inter-event time) and NCD(m,30) (Normalized Cumulative Depth). These indicators, normalized by their past maxima, act as precursor signals of impending significant EQs.
All major EQs—including the 1995 Kobe M7.4, 2011 Tohoku M9.0, 2024 Noto M7.6, and 2025 Kamchatka M8.8 (just outside the monitoring region)—were preceded by NCI and NCD peaks one to two days (and in some cases, hours) before rupture (Fig. 1). This consistency echoes the empirical Accelerated Moment Release (AMR) pattern of increasing seismic release before large EQs (Bufe & Varnes, 1993). While rupture localization is diffuse at the broad scale, the PW framework enables refinement to about 4° × 4° and 1° × 1° meshes through scale-dependent parallel analysis, identifying potential epicentral areas.
Just as meteorologists track typhoons and hurricanes, PWs enable real-time monitoring of seismic conditions. They detect pre-rupture Newtonian oscillations embedded within noisy seismic time-series data extracted from catalogs—signals often missed by traditional methods. This capability represents a paradigm shift in seismic forecasting. For subduction zones such as the Nankai Trough and Cascadia, early detection based on Newtonian physics could significantly reduce disaster impacts and save lives (Takeda, 2015; JP Patent, 2021, 2022a, 2022b, 2023; arXiv, 2024; YouTube, 2025; manuscript in preparation for AGU Advances and arXiv).
Deformation in the brittle crust arises from two sources: direct deformation in the B layer and creep-deformation transfer from the D layer (Zoback & Zoback, 2002). GNSS coseismic strain observations suggest that 80–90% of effective B-layer deformation is direct brittle deformation, while 10–20% reflects ductile creep-deformation transfer. This upward creep-deformation-induced stress accumulation across the D–B transition zone leads to slow, deterministic stress accumulation at the base of the B layer. Such creep-deformation-induced stress accumulation carries physically predictable information about the genesis process and strain-energy accumulation and release cycles of significant earthquakes across subduction zones.
For real-time monitoring, we analyzed seismic catalog data (NIED, 2025) across the Japan subduction zones (16°–52° N, 116°–156° E, 36° × 40° mesh) using Physical Wavelet (PW)-based methods. Seismic time-series data were constructed from events with magnitude >= Mc (=3–4), forming d(c,m) series where c represents latitude (LAT), longitude (LON), depth (DEP), inter-event time (INT), and magnitude (MAG). Here, m denotes the chronological order of events (time). Each series yields three independent active variables with 15–25% stochastic fluctuations, resembling Brownian motion; these variables represent the three principal stress components.
Each earthquake (EQ) is modeled as a virtual EQ particle of unit mass, with successive events tracing an evolving trajectory in stress-sensitive space. Displacement, velocity, and acceleration are extracted as Newtonian quantities by PW operators applied to otherwise stochastic time series. From cumulative INTs and DEPs over 30 consecutive EQs (typically 20–70) with M >= Mc (=4), we defined two normalized strain-energy density indicators: NCI(m,30) (Normalized Cumulative Inter-event time) and NCD(m,30) (Normalized Cumulative Depth). These indicators, normalized by their past maxima, act as precursor signals of impending significant EQs.
All major EQs—including the 1995 Kobe M7.4, 2011 Tohoku M9.0, 2024 Noto M7.6, and 2025 Kamchatka M8.8 (just outside the monitoring region)—were preceded by NCI and NCD peaks one to two days (and in some cases, hours) before rupture (Fig. 1). This consistency echoes the empirical Accelerated Moment Release (AMR) pattern of increasing seismic release before large EQs (Bufe & Varnes, 1993). While rupture localization is diffuse at the broad scale, the PW framework enables refinement to about 4° × 4° and 1° × 1° meshes through scale-dependent parallel analysis, identifying potential epicentral areas.
Just as meteorologists track typhoons and hurricanes, PWs enable real-time monitoring of seismic conditions. They detect pre-rupture Newtonian oscillations embedded within noisy seismic time-series data extracted from catalogs—signals often missed by traditional methods. This capability represents a paradigm shift in seismic forecasting. For subduction zones such as the Nankai Trough and Cascadia, early detection based on Newtonian physics could significantly reduce disaster impacts and save lives (Takeda, 2015; JP Patent, 2021, 2022a, 2022b, 2023; arXiv, 2024; YouTube, 2025; manuscript in preparation for AGU Advances and arXiv).
