Research Methodology

Historical Research Data

The principal Greece research dataset in the 2024 published work was the National Observatory of Athens (NOA) catalogue: earthquake events in the Greek geographic region with magnitude greater than 2.0, spanning 1950 through 2024.

The USGS ANSS Comprehensive Earthquake Catalog (ComCat) was co-utilized as a supplementary global public catalogue for proof-of-concept work and to improve catalogue completeness and minimise loss in historical accuracy metrics. It does not replace NOA as the principal Greece dataset.

Earthquake Potential Score (EPS)

The Earthquake Potential Score is a numerical index ranging from 0 to 100 that reflects how close a specific location is to experiencing its next significant earthquake.

• Calculated using natural time intervals between earthquakes
• Higher scores indicate increased likelihood of seismic activity
• Updated continuously as new seismic data arrives
• Provides quantifiable risk assessment for each forecast region

Natural Time

Natural time is a concept introduced by Varotsos et al. in which event counting serves as a unit of time instead of clock time. It allows seismic sequences to be examined as event-based progression.

• Event-based progression: The number of small earthquakes measures stress and strain accumulation between large earthquakes in a defined geographic area.
• No decluttering required: Natural time can be considered whether aftershocks, background seismicity or both contribute.
• Earthquake cycle definition: The method considers recurring large earthquakes in a broad active region containing numerous faults.
• Entropy dynamics: Natural-time analysis examines the dynamical evolution of the seismic system.

Current S.A.F. Architecture

The current Seismikon Algorithmic Framework combines LSTM Neural Networks and Graph Neural Networks (GNNs) to examine temporal and spatial relationships in monitored seismic data. This is a current framework development, not a claim about the configuration of the 2024 publication.

• Spatio-temporal learning: models examine correlations among events across locations and times.
• Input features: sliding-window outputs can include natural-time statistics, seismic energy release, magnitude distributions and spatial parameters.
• Current predictive output: a five-element (5-tuple) output of longitude, latitude, focal depth, magnitude and time.
• Historical distinction: the 2024 paper described a four-parameter output—coordinates, focal depth and magnitude—with time-frame estimation discussed separately.

Current Dual Sliding Window Technique

The present S.A.F. uses a dual Sliding Window (SW) technique: two windows move inversely along the same time axis to increase dynamic convolutional sensitivity. This is a current framework development.

• Temporal segmentation: seismic catalogues are divided into overlapping time-scaled windows.
• Feature extraction: windows can contain earthquake counts, magnitude measures, seismic energy release and spatial clustering metrics.
• Multi-scale patterns: window sizes can examine shorter-term signals and longer-term seismic trends.

Published Research Results

Retrospective research The 2024 results are retrospective research findings, not live, forward-issued predictions. For prospective (real-time) performance, see the Track Record page.

The published research also investigated occurrence time-frame estimation; time was not one of the four parameters in its reported ≥98% metric. California coverage is magnitude-specific: M≥6 since 1932 and M≥4.5 since 1997.

Current S.A.F. Data Flow