Skip to content
RivuletCheck

Citizen science · urban freshwater

The stream at the end of your street has a health record.

Rivulet turns what residents notice about urban water into scientifically grounded assessments — with the uncertainty stated honestly, and every number traceable to a published method.

See the mapRecord an observation

19

real reports from residents, across 10 streams

4,864

stream sections mapped and ready to check, across 6 cities

Most of that map is waiting for a first report — that gap is the point: it shows exactly where a resident’s five-minute check matters most.

The problem

Nobody can say how much to trust what residents see.

Official monitoring of urban streams is sparse in both space and time. Residents see these waters every day, but what they notice rarely reaches anyone who can act on it — and when it does, nobody can say how much to trust it.

Method

How it works

  1. 01

    You photograph the water.

    We derive a Forel–Ule colour index from the image — a scale in scientific use since the 1890s.

  2. 02

    We weigh the evidence.

    Observations are combined in a Bayesian model that reports a range, not a falsely precise number.

  3. 03

    The city gets an answer.

    Results are shown as indicative status classes with an honest range, and exported as FHIR shaped on the OneAquaHealth implementation guide.

Rigour

There is real science behind this

  • Forel–Ule water colour scale, estimated on the phone with a simplified WACODI-style colour conversion (no camera calibration)
  • Five indicative status classes named after the EU Water Framework Directive — not an official WFD assessment
  • Macroinvertebrate sensitivity set on the BMWP 1–10 family scale — a simplified proxy, not the BMWP protocol
  • Beta–Bernoulli conjugate updating with 90% credible intervals
  • HL7 Europe OneAquaHealth FHIR Implementation Guide
  • Copernicus Sentinel-2 satellite imagery as an independent, coarse cross-check (uncalibrated proxy, not the published hue-angle calibration)

These are proxy estimates from citizen observations, not laboratory measurements. We say so everywhere the numbers appear.

Standards

Open by construction

Every assessment is exportable as FHIR declaring the consortium’s own implementation guide profiles, so this data can flow into the systems that already exist rather than sitting in another silo. Signs residents report travel under a separate, clearly labelled Rivulet code system rather than borrowing laboratory codes.

See the open data