Southern Ontario Shreds

Live whitewater river levels for Ontario paddlers — the Elora Gorge, the Irvine, the Credit, the Madawaska, the Petawawa, the Ottawa, and friends. One page that answers "is it running?" before you drive.

Live site: https://titaniumbones.github.io/so-shreds/

This is a ground-up rewrite of the old river-levels / react-river-experiments projects on a modern stack.

How it works

• Stack: Vite + React 19 + TypeScript. No backend, no proxy — a fully static site.

• Data: every gauge is a federal Water Survey of Canada station, read live in the browser from the ECCC GeoMet OGC API (hydrometric-realtime collection), which serves ~30 days of provisional readings with CORS enabled. The old site depended on a self-hosted CORS proxy and per-conservation-authority scrapers (GRCA/CVC KiWIS); those are gone.

• Runnability bands: each river carries a set of flow ranges (low / shoulder / good / high, in m³/s) drawn as a colored strip with a needle at the current reading, and as washes behind the discharge chart. Thresholds come from the old site's data plus paddler folklore — they are advisory, not safety information.

• Flow prediction (experimental): river pages show a dashed 7-day forecast — a small per-basin rainfall-runoff model (snow + soil store + two linear reservoirs) calibrated offline against decades of daily flows and ERA5 weather, then re-run in the browser against the live Open-Meteo forecast and anchored to the latest gauge reading. On natural-flow rivers it clearly beats "tomorrow = today" at 3–7 day leads; on dam-controlled rivers it can't know what the operator will do, so Shand and Gull don't get forecasts. Full methodology and validation numbers in model/README.md.

• Run journal: every river page has a journal. My runs are private — stored in your browser (localStorage), with the level auto-captured from the live gauge; any entry can be shared as a self-contained link (the entry travels inside the URL, no server). Community reports are GitHub Discussions — one thread per river — posted with your GitHub account; the deploy workflow snapshots comments into reports.json and re-runs on every new comment, so reports appear on the site within minutes. The site itself never touches credentials.

• Virtual gauges: when a physical gauge dies at the source (the Irvine's only station has been down since 2026-07-11), the calibrated model runs unanchored on observed weather and the site shows a clearly-labeled ~modelled estimate instead of a blank.

Development

npm install
npm run dev      # local dev server
npm run build    # static build in dist/

Deploys automatically to GitHub Pages via Actions on every push to main.

Adding a river

Add an entry to src/data/rivers.ts (station number, bands, put-in/take-out coordinates) and optionally a markdown description at public/descriptions/<slug>.md. Find station numbers at wateroffice.ec.gc.ca or by querying https://api.weather.gc.ca/collections/hydrometric-stations/items?PROV_TERR_STATE_LOC=ON.

Data attribution

Hydrometric data: Environment and Climate Change Canada / Water Survey of Canada, via GeoMet-OGC-API. Contains information licensed under the Open Government Licence – Canada. Readings are provisional and subject to revision.