Unify heterogeneous Etna signals without hiding quality, degradation or the distinction between activity nowcast and VONA targets.
3 · What it produces
Activity Index and band, data/sensor health, 1d/5d VONA forecast, prospective verification and adaptive learner in shadow mode.
The cards expose multiple views of the same workflow so result, method, evidence and history remain distinct.
4 · How it works
Bounded acquisition + data health + quality-aware fusion + prospective ledger
1FDSN/seismicity, FIRMS/thermal, cloud data, optional multi-station tremor and event history; freshness, completeness and quality per modality
2quality-adjusted fusion; Activity Index/band; 1h/6h escalation probabilities enabled only after sufficient reconciled evidence; VONA forecast reuses the validated Hawkes/modality gate
3immutable issue ledger, reconciliation sidecars, prospective calibration, hysteresis/cooldown and adaptive learner kept in shadow before gates
4Persistence, rendering and governed publishing.
5 · Data and signals
| Element | Detail |
|---|---|
| Data and signals | FDSN/seismicity, FIRMS/thermal, cloud data, optional multi-station tremor and event history; freshness, completeness and quality per modality |
| Industrial transferability | sensor fusion, condition monitoring, anomaly evidence and governed communication |
| Research | prospective calibration, modality ablation, drift and comparison with strong baselines |
| Interpretation condition | Activity Index, escalation and VONA ash-positive are distinct targets; no output replaces official sources |
quality-adjusted fusion; Activity Index/band; 1h/6h escalation probabilities enabled only after sufficient reconciled evidence; VONA forecast reuses the validated Hawkes/modality gate
immutable issue ledger, reconciliation sidecars, prospective calibration, hysteresis/cooldown and adaptive learner kept in shadow before gates
Python, pandas/NumPy, SciPy/scikit-learn, Matplotlib/Pillow, FDSN/HTTP, GitHub Actions, AT Protocol
mistaking activity nowcast for eruption prediction or VONA ash-positive; sensor degradation/absence and insufficient samples for escalation calibration
Workflow technical dossier
Methods, checks, stack and transferability conditions.
Data and signals
FDSN/seismicity, FIRMS/thermal, cloud data, optional multi-station tremor and event history; freshness, completeness and quality per modality
Inference / calculation
quality-adjusted fusion; Activity Index/band; 1h/6h escalation probabilities enabled only after sufficient reconciled evidence; VONA forecast reuses the validated Hawkes/modality gate
Validation
immutable issue ledger, reconciliation sidecars, prospective calibration, hysteresis/cooldown and adaptive learner kept in shadow before gates
Stack
Python, pandas/NumPy, SciPy/scikit-learn, Matplotlib/Pillow, FDSN/HTTP, GitHub Actions, AT Protocol
Limitations
mistaking activity nowcast for eruption prediction or VONA ash-positive; sensor degradation/absence and insufficient samples for escalation calibration
See recent outputs
The Publications page collects local previews and, on request, the public feed.