Multi-source data layers

A layer is one dataset drawn on the shared map: earthquake epicentres, volcano alert levels, fire hotspots, food prices by province, vessel positions, weather warnings. The platform ships 85 or more of them, organised by pillar and each independently toggleable, so an operator sees only the combination their work needs.

Because the layers share one map, one timeline and one time filter, they combine. Hazard layers, economic indicators and infrastructure can be read on the same view rather than compared across screenshots taken at different moments.

• Hazards — seismic, volcanic, ground movement, fire and air-quality layers.
• Maritime — vessel positions, ports, waterway boundaries and passage risk.
• Economic — prices, trade flows, markets and development indicators.
• Incidents and media — public-order incidents, demonstrations and aggregated coverage.
• Reference — administrative boundaries, zones and geometry to read the data against.

The catalogue is searchable and filters by pillar, layers can be shown or hidden in bulk, and any layer can be marked as a favourite so a working set survives a session. Volume layers carry sub-layers — related views grouped under one heading — so a dataset with severity bands can be read one band at a time.

Presentation is per layer as well. Opacity and time range are set on the layer itself, so a dense background dataset can be held back while the layer being worked on stays prominent, and a whole catalogue can be trimmed down to the handful of layers a shift actually watches.

• Built-in layers — the catalogue that ships with the platform, sourced and maintained centrally.
• Custom layers — your own imports, kept apart from the built-in catalogue.
• Public and shared layers — layers published for a wider set of users.
• Favourites and settings — per-user organisation of a large catalogue.
• Sub-layers — severity bands and related views grouped under one parent layer.

Custom data enters without a developer. An ArcGIS FeatureServer or MapServer URL is validated against the live service and its metadata fetched before the layer is accepted, so a broken endpoint is rejected at the point of entry rather than discovered later; GeoJSON can be supplied by URL, file or paste, and its structure and bounds are checked on import.

Field input belongs to the same timeline as the machine feeds. Operator reports submitted in the field are ingested as data points, so a survey observation and a satellite detection appear in one playback instead of in two systems that never meet.

• ArcGIS — FeatureServer and MapServer services checked against the live endpoint before acceptance.
• GeoJSON — supplied by URL, file or paste, with structure and bounds validated.
• Own points and perimeters — places that matter to the operator, drawn on the map.
• Field reports — operator submissions folded into the same timeline as the automated feeds.

Behind the catalogue is a source registry: every data source with its schedule, timeout and retry policy, a manual trigger, and the run history of each attempt with the next scheduled run time. Ingestion runs as a separate worker process, so a slow or failing feed cannot degrade the interface.

Freshness is therefore visible rather than assumed. When a layer looks thin, the run log says whether the last attempt succeeded, failed, or has not run yet, which is the difference between a quiet period and a broken pipeline.

• 38 pipelines — each source has its own job, so one failure does not stop the others.
• Schedules — intervals set per source, with a manual trigger for an on-demand refresh.
• Run log — every attempt recorded with its outcome, so a gap in a layer can be explained.
• Worker isolation — ingestion runs outside the interface process.

Every dated layer answers to the same clock, which is what makes comparison possible. A single timeline scrubs across all of them, and a range filter narrows the view to the last hour, the last day or the last week without changing which layers are switched on.

That shared clock also means a layer's history is not a separate feature bolted on later. Playback, density views and province-level aggregation all read the same data the current view does, so an operator never has to reconcile two versions of the same event.

One map is only half the benefit; one clock is what makes two layers comparable in the first place.

Every layer inherits the cadence and the coverage of the source behind it, and the platform shows both rather than hiding them: an hourly feed and a daily one will not look equally fresh, and the registry is where that difference is visible. Operators are told what to expect from each layer rather than left to discover it in the middle of an incident.

Access is scoped to the operator's role, so a layer that is published for a wider set of users can still be held back from people who should not see it. Custom imports belong to their owner, and shared layers are published deliberately rather than by default.