Plain-language summary
- What it solves
- Official aviation advisories for Etna are public but hard to read outside the field.
- Who it can serve
- Anyone wanting to follow the advisories in plain form, and anyone studying automated outreach.
- What it produces
- An advisory card, 12-hour trajectories, data quality notes and posts in Italian and English.
To automate collection, checks, calculation and publication while keeping a hard boundary between the authoritative source and non-operational simulation.
3 · What it produces
Official VONA relay and educational scenario remain visually and logically separate.
When a new positive VONA appears, the workflow blocks the educational simulation and can relay the official notice through independent IT/EN root posts. Under CLEAR status, the pipeline can produce four educational cards covering trajectories, proximity, quality and limitations.
4 · How it works
From authoritative source to the gate, then to the appropriate model.
The trajectory moves a point hourly using wind speed and direction at the selected pressure level. Alignment is the minimum Haversine distance to simplified transport geometries; its score is heuristic, not a probability.
5 · Data used
| Source | Information | Use in calculation | Control |
|---|---|---|---|
| INGV VONA | official bulletin and POSITIVE/CLEAR/UNKNOWN status | source gate and event relay | fail closed if the authoritative source cannot be verified |
| Open-Meteo | forecast wind at 700 and 500 hPa | hourly vectors for the 12h trajectory | NaN/completeness checks and live-data validity |
| NASA FIRMS | thermal anomalies and FRP | optional context | never treated as proof of ash |
| NASA GIBS VIIRS | true colour + thermal anomalies | visual context | graceful fallback; not ash evidence |
| Transport assets | simplified airports, roads, rail and ports | minimum distance and alignment band | geometry screening, not plume width or impact |
Advection and distance
Hourly 12h step; wind vector converted to Δlat/Δlon; Haversine distance and broad alignment thresholds.
Isolation Forest on winds
Isolation Forest provides only a TYPICAL/UNUSUAL percentile for available wind context: it does not detect ash or predict eruptions.
Fail-closed and idempotent
GitHub Actions orchestrates scheduled/manual runs; positive or unknown official state blocks simulation. Run keys and monitor state prevent intentional duplicates.
Not a warning system
No estimate of concentration, deposition, visibility, engine risk, closure or service disruption.
Workflow technical dossier
Methods, validation and implementation stack.
Runtime and packages
- pandas and NumPy for data and calculations.
- scikit-learn / IsolationForest only for wind anomaly context.
- Matplotlib and Pillow for cards; BeautifulSoup and pypdf for controlled parsing.
- pydantic and requests for config/HTTP; GitHub Actions and AT Protocol/XRPC for orchestration/publication.
Geometry
- Trajectories at about 3 km (700 hPa) and 5.6 km (500 hPa) in the current configuration.
- Asset polylines are densified before minimum-distance calculation.
- ELEVATED/MODERATE/LOW/NO MATERIAL ALIGNMENT bands are distance-based heuristics.
Governance
- Mandatory disclosure on cards/captions and traced provenance.
- FIRMS/GIBS are thermal/visual context, never ash evidence.
- VONA monitor every three hours plus fallback within production workflow.
Publishing
- Independent IT/EN root posts and up to four images per post.
- Event relay is deduplicated and shares concurrency with weekly production.
See recent outputs
The Publications page can show multiple images from the same post or preview, with gallery navigation.
Potential transferability
Have a problem similar to Etna Sentinel?
A first discussion can start from the objective, available data and success criterion without sending confidential information. A result observed here is not assumed to transfer automatically to another context.