Methodology

How GlobeWebcams discovers, checks, organizes and publishes webcam information.

1. Source discovery and rights review

GlobeWebcams starts from official or provider-controlled inventories where possible. Public reachability alone is not considered reuse permission. Source licensing, attribution requirements and allowed forms of use are tracked independently from the camera metadata.

2. Sites and views are separate

A physical camera installation is modeled as a camera site. Directional presets or separate views can belong to the same site. This prevents multi-view cameras from being presented as unrelated locations when the source provides enough evidence to group them.

3. Geographic enrichment

Provider coordinates are used when available. Place and country labels can be enriched from geographic reference data so cameras can be browsed by location. A nearby place label is discovery context and does not imply that a camera is physically located at the center of that place.

First-level region names and selected named natural features are drawn from a dated, country-specific GeoNames extract, licensed under CC BY 4.0. The site stores the source record and code behind each published fact. Provider location metadata is used separately. Point-to-point proximity is labeled as nearby, not as proof that a camera lies inside a feature. No terrain or climate classification is inferred from a camera name or image.

4. Health evidence is persistent

A successful request by itself is not enough to call a camera live. Repeated observations are used to distinguish confirmed, probable, unknown and stale states. Under the current default policy, an unchanged frame becomes suspected stale only after at least 12 hours and at least 6 matching observations. Confirmed stale normally requires at least 48 hours, at least 12 matching observations, and reliable provider freshness evidence that has also remained old. When a provider's HTTP freshness metadata is known to be unreliable, GlobeWebcams ignores that metadata and uses a stricter source-specific repeated-content rule instead. A previously live camera is not removed after one transient request failure; the current default requires 3 consecutive failures before it becomes offline. Suspected-stale cameras may remain visible on the map but are not eligible for indexing; confirmed-stale cameras are not published. Health checking does not create an image archive.

5. Source inventory changes use a grace period

For providers that publish a complete snapshot, a camera missing from one refresh is not immediately retired. The current default keeps the previous record for 48 hours before absence from repeated complete snapshots can retire it. A complete-snapshot refresh that is implausibly smaller than the active baseline is rejected instead of being treated as mass camera removal. Sources identified as partial or incremental do not use absence from a refresh as an automatic retirement signal.

6. Map visibility and indexing are different

Publication gates consider source release status, rights, health evidence, duplicate review and page quality. Experimental, candidate, key-gated or rights-pending sources remain unpublished until their source status is explicitly promoted to a production-ready state. Among production-ready sources, cameras can still be visible for map discovery without automatically becoming indexable pages or entering the sitemap.