Platform
Search
Most sensor data is public and almost none of it is findable. Every network publishes in its own format, under its own identifiers, on its own host. Search is the layer that makes them one searchable surface.
What you can search by
- Place — city, region or country. Queries resolve against a gazetteer, so "water level germany" returns German river gauges rather than documents containing those words.
- Network — filter to a specific operator, from national meteorological services to community projects.
- Measurement — temperature, PM2.5, water level, occupancy, and dozens more, normalised across networks that each name them differently.
- Coordinates and radius — everything within a distance of a point.
Typo tolerance is not a nicety
Station names are a mess of transliterations, diacritics and local spellings. A search engine that only does exact matching is unusable against them: someone looking for Århus will type Aarhus, or Arhus, or mistype it entirely.
Search is built on a typo-tolerant index, so near-misses still land. "Aarhsu" resolves. "Munchen" finds München. That single property does more for real-world usability than any amount of ranking tuning.
Ranking favours live data
A station that exists on paper and a station reporting right now are not equally useful. Results are ranked with a has-data signal, so sensors carrying a recent reading sort above catalogue entries that have never reported. Registry records stay searchable — you can still find a station by name — they simply do not crowd out live instruments.