Hazard monitoring layers

The hazard group is the platform's environmental core: real-time and historical earthquakes, volcanoes with their current alert level, ground movement, fire hotspots, air quality and weather warnings — each with its own symbology, so severity reads off the map without opening a table. These are the layers the rest of the platform reasons over, which is why they are maintained as production feeds rather than as occasional downloads.

Seismic activity is ingested from two independent feeds, which means a significant event can be cross-checked against a second source instead of being trusted from a single report. Everything in the group carries a date, which is what ties it to the shared timeline.

• Earthquakes — real-time and catalogue views, magnitude-scaled, with hexbin density for regional patterns.
• Volcanoes — all 127 monitored volcanoes with alert status shown as an animated status marker.
• Landslides — ground-movement events, read alongside weather and terrain.
• Fire hotspots — satellite detections with a confidence band, and a separate view restricted to hotspots inside forest boundaries.
• Weather and air — active warnings with their affected areas, and air-quality stations with their category.

Symbology carries meaning before a single popup is opened. Marker size follows magnitude, so the largest events are visible at national zoom, and real-time alerts are drawn with concentric rings that hold attention on a picture that is moving.

Density has its own views — hexagonal grids for regional patterns — and clustering can be restricted to severe points only. That last choice matters more than it looks: it means a cluster on the map represents severity rather than the accident of how many points happen to be in one place.

• Magnitude scaling — marker size follows event strength.
• Alert rings — visible emphasis for real-time seismic alerts.
• Hexbin density — regional patterns rather than individual points.
• Severity-aware clustering — only high-severity points cluster, so density means something.
• Forest context — a hotspot inside protected forest is shown as the different emergency it is.

Every dated layer answers to the same timeline: a 24-hour window with playback and speed control, and range filters for the last hour, day, week or the full history. Playback is how a sequence becomes legible — the aftershock train, the fire front, the weather system that crossed three provinces.

Clicking a province aggregates what is inside it, so a regional question is answered with the layers already on the map rather than a fresh export. The same geometry is used for every layer, so two sources can be compared without one of them being rounded to a coarser area.

A reading without its history is a number; the timeline is what turns a layer into a situation.

Hazard feeds update continuously and layer state survives a refresh, so an operator watching an evolving event does not lose the configuration they built around it. Where a feed lags, the layer shows what it has and when it last updated.

Sources are ingested on their own schedules with their own retries, and a failure in one feed does not take the others off the map. That separation is deliberate: during a real event, the layers that are still current are exactly the ones that matter most.

• Independent schedules — each feed refreshes at its own interval.
• Retries — a failed fetch is retried rather than abandoned for the day.
• Last-updated state — a lagging layer says so instead of quietly showing old data.
• Stable configuration — toggles, opacity and time range survive a refresh.

These layers are not an end in themselves. They are the inputs the rest of the platform reasons over, which is why the same reading can appear as a map, an alert and a written paragraph on the same day.

That makes the hazard group the common reference point of the platform: a rule fires on a hazard metric, an event becomes an entity in the graph, and the report generator summarises the same activity in prose for whoever was not watching the screen. One reading therefore arrives in whichever form the work in front of you needs.

• Rules — alert thresholds and anomaly baselines are evaluated against hazard metrics.
• Graph — hazard events are entities, so an event links to the places and organisations around it.
• Reports — the automated situation report summarises hazard activity for the period it covers.
• Workflow — an event on the map can be carried into the response board that handles it.

What arrives is what the sources publish. A feed that reports on a delay, that revises earlier figures or that covers some regions more densely than others will show exactly that on the map, and the layer's own timestamp is the honest answer to how current it is.

Nothing is interpolated to fill a gap. Where a sensor network is sparse, the map shows sparsity, because a smoothed hazard layer would look complete while being wrong in precisely the places where being wrong matters.