Seismic Risk Analysis

Tokyo & Kanto Plain

35.69°N, 139.69°E R = 250 km 103 Yıl Birikim 0.8 Metre (Sagami Trench)

Sagami Trench subduction + İntra-slab gerilimi. Philippine Sea Plate batıya dalar. — Sagami Trench subdüksiyon bölgesi + Tokyo Metropol fay zonu. Kayma hızı 40–50 mm/yr (Philippine Sea Plate); son büyük olay: 1923 Büyük Kanto M7.9.

Instant Risk Assessment
5 Magnitude Band Forecast
Regional Model v2
Bölgeye Özel Band Tahminleri

Koordinata en yakın önceden eğitilmiş bölgesel model kullanılır. Her band ayrı stratejiyle eğitilmiştir (zor negatifler, temporal validation).

Yükleniyor…
Physical Framework
Seismic Cycle Summary
Seismic Gap
103 Yıl Birikim
Son kırılmadan beri
Slip Deficit
0.8 Metre (Sagami Trench)
Birikmiş plastik deformasyon
Slip Rate
40–50 mm/yr (Philippine Sea Plate)
Jeodezik GNSS ölçümü
Max. Moment
M≥8.0
Segment potansiyeli
b-Value Trend
—
Model çıktısından
Dominant Band
—
En yüksek risk skoru
Tidal & Lunar Cycle
Celestial Alignment Window
Historical Seismic Activity
Magnitude Band Catalog Distribution

Kaynak: USGS, AFAD, ISC Bulletin (1900–2026) — Tokyo & Kanto Plain, R=250 km. Değerler 100 yıllık normalize sıklığı temsil eder.

Spatial Risk Distribution
Region Risk Heatmap

250 km yarıçap içinde 12×12 ızgara üzerinde ML risk skoru. Yeşil → düşük, sarı → orta, kırmızı → yüksek.

Detailed Risk Analysis Report
High-resolution grid, fault proximity scores, and 5-year probability forecasts.
Get Full Report
Regional Geology
Fault System Details
Fault systemSagami Trench subdüksiyon bölgesi + Tokyo Metropol fay zonu
Fault typeSubdüksiyon + içslab
Slip rate40–50 mm/yr (Philippine Sea Plate)
Last major event1923 Büyük Kanto M7.9
Segment potentialM≥8.0
Seismic Gap103 Yıl Birikim
Slip Deficit0.8 Metre (Sagami Trench)
ML Model Metrics
Band-Level ROC-AUC Accuracy
M3–4 —
M4–5 (RandomForestClassifier) 0.860
M5–6 (GradientBoostingClassifier) 0.955
M6–7 (GradientBoostingClassifier) 0.997
M7+ (ExtraTreesClassifier) 0.993
Model Validation
Backtest & Forecast Accuracy

Walk-forward backtest from 1990 to present — at every time slice the model is trained only on past data and tested on future events.

—
Avg. ROC-AUC
—
Brier Score
—
ECE (Calibration)
—
Test Windows
—
Skill Score (Poisson)
Prospective Forecast Accuracy

Every forecast is stored and verified against the USGS catalog when its window closes. Real performance — not retrospective.

Loading accuracy…
Prospective Validation
Past Forecasts (Last 60 Days)

Forecasts issued by the system and their outcomes. Each forecast window is automatically matched against the USGS catalog when it closes.

Loading accuracy…
Forecasts are automatically saved and verified against the USGS catalog.