Researcher Verification Kit
Independent, reproducible verification of the BRETT v4.2.0 earthquake prediction engine. Run our test harness to validate accuracy across 33 seismic zones against real USGS data.
Download Test Harness
The Python verification suite tests across all major seismic zones, calculates accuracy at three tolerance levels, and generates institutional reports with confidence intervals.
Quick Start
# Install dependencies pip install -r requirements.txt # Quick test (5 zones, 2 time periods) python3 regenisis_verification.py --quick # Full verification (33 zones, all periods) python3 regenisis_verification.py # Report from existing data python3 regenisis_verification.py --report-only
Public API Documentation
Historical Backtesting API
GETPublichttps://earthpulse.abacusai.app/api/predictions/backtestParameters
latโ
Latitude (-90 to 90)lngโ
Longitude (-180 to 180)startโ
Start date (YYYY-MM-DD)endโ
End date (YYYY-MM-DD, must be past)nameโLocation name labelradiusKmโSearch radius (default: 500, max: 1000)minMagโMinimum magnitude (default: 4.0)toleranceโMagnitude tolerance (default: 1.5)Example
curl "https://earthpulse.abacusai.app/api/predictions/backtest ?lat=35.7&lng=139.7 &start=2023-01-01 &end=2024-01-01 &name=Japan"
Response
{
"location": { "name": "Japan" },
"accuracy_tolerance_1_5": 87.3,
"accuracy_tolerance_1_0": 71.2,
"accuracy_tolerance_0_5": 48.6,
"total_events_tested": 127,
"data_source": "USGS"
}Prediction Ledger API
GETPublichttps://earthpulse.abacusai.app/api/predictions/ledgerParameters
statusPENDING, CONFIRMED, REFUTED, NO_EVENT, ALLformatjson (default) or csvregionFilter by region name (partial match)pagePage number (default: 1)limitRecords per page (max: 200)sortcreatedAt, verifiedAt, magnitude, accuracyorderasc or descExamples
# All predictions (JSON) curl https://earthpulse.abacusai.app/api/predictions/ledger # Confirmed only curl https://earthpulse.abacusai.app/api/predictions/ledger?status=CONFIRMED # Download as CSV curl https://earthpulse.abacusai.app/api/predictions/ledger?format=csv
Unified Volcanic-Earthquake Engine
Two-State Eruption Model โ Resonance ร Trigger Activation
๐ง The Core Insight (Nicolas's Two-State Model)
Volcanoes don't erupt on a schedule. They erupt when an earthquake shatters the conduit. This engine monitors BOTH conditions in real-time:
Unified probability combines resonance output with seismic trigger analysis
through a proprietary activation function.
Status transitions:
DORMANT โ HOT_MONITORING โ IMMINENT โ ERUPTION_LIKELY
โ โ
โโโ ELEVATED โโโโโโโโโโโโโโโ (72hr window expires)๐๏ธ Volcano-Specific Fragility Curves NEW
Not all volcanoes are equal. Each has its own "fragility curve" based on chamber depth, crust type, and eruptive history. The threshold is now auto-calculated per volcano.
Threshold is auto-calculated per volcano based on: - Chamber depth - Crust type (oceanic, continental, caldera, etc.) - Eruptive history frequency Lower threshold = more fragile volcano = easier to trigger
๐ก Use &threshold=XX to manually override the auto-calculated threshold for custom analysis. Omit the parameter to use the volcano's natural fragility curve. 20 major volcanoes profiled.
lat*Latitude (-90 to 90)lng*Longitude (-180 to 180)nameVolcano name (enables auto-threshold from fragility profile)thresholdManual override (auto-calculated per volcano if omitted)lookback_daysEarthquake lookback days (1-90, default: 30)radius_kmEarthquake search radius (50-500, default: 300)daysForecast days (1-30, default: 7)Example Requests
# Marapi โ auto-threshold ~38 (FRAGILE) curl "https://earthpulse.abacusai.app/api/volcanic/unified?lat=-0.38&lng=100.47&name=Marapi" # Kilauea โ auto-threshold ~28 (VERY FRAGILE) curl "https://earthpulse.abacusai.app/api/volcanic/unified?lat=19.42&lng=-155.29&name=Kilauea" # Yellowstone โ auto-threshold ~62 (VERY STURDY) curl "https://earthpulse.abacusai.app/api/volcanic/unified?lat=44.43&lng=-110.67&name=Yellowstone" # Etna โ manual override to 35 curl "https://earthpulse.abacusai.app/api/volcanic/unified?lat=37.75&lng=14.99&name=Etna&threshold=35"
Eruption Status States
Response Includes
Volcanic Engine API
Live volcanic forecasting with full calculation breakdown + earthquake trigger analysis
Live Volcanic Forecast
Returns volcanic eruption probability forecast with earthquake destabilization analysis from real USGS data.
lat*Latitude (-90 to 90)lng*Longitude (-180 to 180)nameVolcano/location namedaysForecast days (1-90, default: 21)Example Requests
# Marapi (Indonesia) โ currently high activity curl "https://earthpulse.abacusai.app/api/volcanic/engine?lat=-0.38&lng=100.47&name=Marapi" # Kilauea (Hawaii) โ 30 day forecast curl "https://earthpulse.abacusai.app/api/volcanic/engine?lat=19.42&lng=-155.29&name=Kilauea&days=30" # Etna (Italy) curl "https://earthpulse.abacusai.app/api/volcanic/engine?lat=37.75&lng=14.99&name=Etna"
Volcanic Historical Backtest
Monthly volcanic predictions vs actual events with earthquake-trigger correlation per month. Tests the theory: eruptions need a seismic trigger to destabilize the chamber.
lat*Latitude (-90 to 90)lng*Longitude (-180 to 180)start*Start date (YYYY-MM-DD)end*End date (YYYY-MM-DD, must be past)nameVolcano/location nameradiusKmSearch radius (50-500, default: 300)Example Requests
# Marapi 2024 backtest curl "https://earthpulse.abacusai.app/api/volcanic/historical?lat=-0.38&lng=100.47&start=2024-01-01&end=2025-01-01&name=Marapi" # Kilauea 2024 curl "https://earthpulse.abacusai.app/api/volcanic/historical?lat=19.42&lng=-155.29&start=2024-01-01&end=2024-12-31&name=Kilauea"
๐ Earthquake Trigger Hypothesis (Nicolas's Theory)
Actual surface eruptions require an earthquake trigger event to destabilize the magma chamber/pipe. Without a seismic trigger, even high-resonance volcanic systems may not erupt. The API calculates destabilization potential for each nearby earthquake using:
Destabilization potential is calculated using: - Seismic energy release (Gutenberg-Richter relation) - Distance attenuation from earthquake to volcano - Depth coupling between earthquake focus and magma chamber
Risk thresholds: NEGLIGIBLE (<10) โ LOW (10-25) โ MODERATE (25-45) โ HIGH (45-70) โ CRITICAL (>70)
Response Includes:
Expected Output
Console Output
============================================================
REGENISIS COSMOLOGY โ Independent Verification Suite
============================================================
Tests: 165 | Zones: 33 | Periods: 5
๐ Japan (35.7, 139.7)
[1/165] Full Year 2023 (2023-01-01 โ 2024-01-01)
โ Events: 127, Acc: 87.3% (ยฑ1.5)
...
============================================================
VERIFICATION SUMMARY
============================================================
Global Accuracy (ยฑ1.5): 87.3%
Global Accuracy (ยฑ1.0): 71.2%
Global Accuracy (ยฑ0.5): 48.6%
Weighted Accuracy: 89.1%
vs Random Baseline: +12.4%
95% CI: ยฑ3.2%
============================================================Generated Files
verification_reports/ โโโ consolidated_report.json โโโ verification_report.md โโโ all_backtest_results.csv โโโ ledger_snapshot_2026-04-10.json regenisis_verification.db (SQLite)
What Gets Validated
Methodology & Transparency
No data leakage โ each prediction uses only prior month's data as context
Three tolerance levels โ ยฑ0.5, ยฑ1.0, ยฑ1.5 magnitude
USGS source of truth โ all verifications link to USGS event pages
Immutable ledger โ predictions timestamped before events, never deleted
Reproducible โ anyone can run the same test harness
Baseline comparison โ improvement calculated over random chance