We Don't Predict Earthquakes. We Calculate Seismic Risk with Millimetric Precision.
Free weekly seismic risk forecasts for every active fault zone. Need the full picture for your exact address? Get a professional 5-page PDF report — fault analysis, 5-year probabilities, and AI risk scores across all magnitude bands — delivered to your inbox for €5.
Classic AI models cry wolf at every minor tremor, inflating their accuracy numbers. Talivio works differently. Our system is deliberately under-confident — it never triggers an alert when the evidence is uncertain.
When our system issues a "Risk > 0.7" warning, seismic activity in the target region is confirmed 91% of the time across 15 walk-forward folds.
Say goodbye to false alarms. Know only the moments that truly require action.
Talivio's performance compared to published earthquake forecasting systems worldwide.
Our AUC values use hard same-region negatives — the most rigorous evaluation protocol.
Values auto-update as models are retrained.
System
Metric
Value
ASS†
Window
Negatives
Validation
ETAS Italy (OEF-Italy)
Area Skill Score
0.7
0.700
1-day
same_region
Prospective
RELM California (CSEP)
Probability Gain vs Random
10
—
5-year
same_region
Prospective
CSEP California (2011–2020)
IGPE vs Benchmark
0
—
5-year
same_region
Prospective
FCN Deep Learning (California)
Area Skill Score
0.882
0.882
15–90 day
same_region
Retrospective
DeVries 2018 (Google Brain)
AUC
0.849
—
static
global_mixed
Retrospective
ETAS Japan (CSEP daily)
CSEP Pass Rate
0.9
—
1-day
same_region
Prospective
EEPAS New Zealand
IGPE
0.64
—
3-month
same_region
Prospective
Talivio v2
AUC + ASS (Molchan)
0.619–0.994
0.218
30-day
Hard same-region
Retrospective + Prospective
† ASS = Area Skill Score (Molchan diagram, normalized [0,1]). FCN Deep Learning=0.882, ETAS Italy=0.70 from published papers.
Talivio ASS computed live from verified prospective forecasts — updates automatically as outcomes are confirmed.
* AUC world ranking based on honest evaluation: hard same-region negatives (model learns "will this fault rupture?" not "is this a fault?").
Most published ML AUC >0.90 use geographically distinct negatives.
8 regions, 127 backtest years. Full methodology →
Architecture
How It Works
Mechanical & Celestial Stress
Fault geometry (strike-slip/normal) and Coulomb stress transfers are mapped to compute physical fracture points.
Epidemic Clustering (ETAS)
Hundreds of thousands of past earthquakes are analysed via Morlet Wavelet transforms and temporal decay to separate background noise from genuine hazard signals.
Spatial Network Awareness
Using Uber's H3 hexagonal architecture, seismic stress transfer between neighbourhoods is analysed with graph-based logic.
Walk-Forward Validation
25 Years of Uninterrupted Validation
Our system has been tested with Walk-forward Validation and a 30-day purge gap protocol to prevent data memorisation.
0.809
Median M4-5 AUC
Across 8 regions, hard same-region negatives
0.994
Best Regional AUC
Highest M4-5 AUC across all regions
127
Backtest Years
Total years of expanding-window validation
0.218
Area Skill Score
Molchan ASS (prospective). FCN=0.882, ETAS Italy=0.70
Location-Specific Models
Location-Specific ML Models
Each region trains its own ML model. Negative samples are drawn from the same tectonic zone — producing honest AUC values (0.62–0.99 range) with no geographic bias.
Istanbul & Marmara
0.619
AUC · M4-5
M3-4
M4-5
M5-6
M6-7
M7+
Kahramanmaras & EAF
0.796
AUC · M4-5
M3-4
M4-5
M5-6
M6-7
M7+
Yedisu Seismic Gap
0.738
AUC · M5-6
M3-4
M4-5
M5-6
M6-7
M7+
Izmir & Aegean
0.702
AUC · M4-5
M3-4
M4-5
M5-6
M6-7
M7+
Cascadia Subduction
0.940
AUC · M3-4
M3-4
M4-5
M5-6
M6-7
M7+
Tokyo Metropolitan
0.809
AUC · M4-5
M3-4
M4-5
M5-6
M6-7
M7+
Nepal-Tibet Himalayan
0.983
AUC · M4-5
M3-4
M4-5
M5-6
M6-7
M7+
San Andreas South
0.854
AUC · M4-5
M3-4
M4-5
M5-6
M6-7
M7+
Zagros Thrust Belt
0.894
AUC · M5-6
M3-4
M4-5
M5-6
M6-7
M7+
NZ Alpine Fault
0.994
AUC · M4-5
M3-4
M4-5
M5-6
M6-7
M7+
Amber band = trained model exists · AUC = expanding-window backtest result (hard same-region negatives)
What's Inside
Your Report: 5-Page Professional PDF
✓AI risk scores across all 5 magnitude bands (M3–4 to M7+)
This system provides statistical probability forecasts. Earthquake time, location and magnitude cannot be precisely predicted scientifically. Probabilities are based on historical patterns and physical models; they do not imply certainty.