About Seismikon
Our mission, the science, and the commitment to transparent earthquake nowcasting
Why Seismikon Was Created
Earthquake prediction has long been considered an unsolved problem - and for classical approaches, that characterisation is largely fair. But a new class of methodology, grounded in natural-time statistical mechanics and trained on decades of seismic catalogues, is producing prospective forecasts that can be evaluated against real outcomes.
Seismikon was created to take that research out of the laboratory and into the world. We believe that if a method can produce predictions that are falsifiable, specific, and independently verifiable, it deserves public scrutiny - not just academic peer review. The best way to earn credibility is to publish predictions before they can be evaluated, publish the evaluation criteria in advance, and then publish the results - including every failure.
That is precisely what Seismikon does.
The Scientific Basis
Natural Time
Instead of measuring time by the clock, natural time counts seismic events - the number of earthquakes that have occurred since some reference point. This event-based representation reveals patterns in the seismic cycle that are invisible in ordinary time.
Earthquake Potential Score
The Earthquake Potential Score (EPS) quantifies how far along the seismic cycle a given region is. A high EPS indicates that the region has accumulated sufficient "seismic readiness" for a significant event. EPS is derived from natural-time analysis of the regional seismic catalogue.
LSTM Neural Networks
Long Short-Term Memory (LSTM) networks are trained on historical seismic sequences to learn the temporal and spatial dependencies that precede significant earthquakes. The networks take EPS and related features as inputs and produce predictions of event location, magnitude, depth, and time.
Prospective Evaluation
Every prediction is evaluated against a predefined set of tolerances: geographic distance, magnitude range, depth, and time window. A prediction is matched only if all four criteria are satisfied simultaneously. The tolerances are published in advance and do not change after the fact.
Peer-Reviewed Research
The methodology implemented in Seismikon is grounded in the published work of Dr. Anastasios N. Bikos (University of Patras, Greece). The primary research paper describing the earthquake nowcasting approach has undergone peer review and is freely available.
Seismic Nowcasting with Machine Learning
Describes the natural-time methodology, the LSTM architecture, the Earthquake Potential Score formulation, and the prospective evaluation framework used in Seismikon. Includes validation results from retrospective and prospective tests on the California seismic catalogue.
Download PDFAdditional publications will be listed on the Research page as they are released.
Commitment to Publishing Failures
The seismic prediction field has historically been plagued by selective reporting - researchers publicising apparent successes and quietly ignoring failures. Seismikon takes the opposite approach.
No post-hoc adjustments, no unpublished test runs, no cherry-picking.
Tolerances for location, magnitude, depth, and time are defined before the prediction is made and do not change.
The full track record - including all unmatched predictions - is available on the Track Record page.
Prediction records and outcome datasets are available for download on the Data Downloads page.
When a matched prediction corresponds to an earthquake with casualties or major damage, all communications reflect the gravity of that event.
Commercial Objective
Seismikon is not purely academic. The platform is designed to be commercially viable, and we believe that commercial pressure - correctly applied - is a useful accountability mechanism: if the predictions are not good enough to be useful in real-world applications, the business fails.
Current commercial applications include:
- Infrastructure risk management - helping operators of critical infrastructure (energy, water, transport) anticipate elevated seismic risk periods
- Insurance underwriting - providing short-term seismic risk signals to complement long-term probabilistic hazard assessments
- Emergency preparedness - supporting public safety organisations in planning and resource pre-positioning
- Scientific research - providing a live, auditable prediction dataset for independent seismological research
Global Applicability
We have set up a public demonstration using California because it has one of the world's most complete, real-time seismic catalogues (via USGS/FDSN) and high seismicity - meaning predictions can be evaluated against real events on a useful timescale. This is a demonstration, not a claim that the system is California-specific.
The underlying methodology is not tied to any single region. Any seismically active area with a sufficiently dense historical catalogue can be modelled using the same approach. Regions under evaluation for commercial coverage include:
[ Global region expansion list - to be published when commercially available ]
Contact commercial@seismikon.com to discuss coverage for a specific region.
Ethical Approach to Communicating Predictions
Earthquake prediction carries unique ethical weight. Done badly, it can cause unnecessary panic, erode public trust in scientific institutions, or - worse - lead to complacency when no prediction is issued. Seismikon takes this responsibility seriously.
We do not issue warnings
Seismikon produces predictions, not warnings. We are not an official government warning system and do not advise evacuation or specific protective actions. All guidance should come from authorised seismic monitoring agencies.
We do not exaggerate certainty
All predictions are probabilistic. We communicate uncertainty explicitly and do not imply that a prediction guarantees an earthquake will or will not occur.
We apply special sensitivity after damaging events
Following earthquakes with casualties or major damage, we do not issue triumphant communications about matched predictions. The human cost of such events takes precedence over any commercial or reputational consideration.
We support independent verification
We publish our evaluation methodology and raw data so that independent researchers can verify - or dispute - our performance claims. We welcome critical engagement from the scientific community.
Meet the Founders
Seismikon is co-founded by Dr. Anastasios N. Bikos (University of Patras, Greece) and Jensen Technologies, combining world-class seismological research with production-grade engineering.