How Nibi Metrics works

Nibi Metrics turns raw gauge readings from official hydrometric networks into something a person standing on a dock can actually use: is the water high or low for today, which way is it moving, and how unusual is this? This page explains where every number on the platform comes from and what we do to it along the way.

Where the data comes from

Every reading on Nibi Metrics originates from a public, government-operated monitoring network:

Water Survey of Canada (WSC). The federal hydrometric network, operated by Environment and Climate Change Canada, provides real-time water levels and discharge for stations across every province and territory. WSC data is the backbone of our Canadian coverage.

CEHQ (Quebec). The Direction de l'expertise hydrique du Québec operates an extensive provincial network. We integrate CEHQ stations alongside WSC coverage in Quebec, published under Quebec's open-data license (CC-BY 4.0).

U.S. Geological Survey (USGS). Our American coverage is built on the USGS National Water Information System — the same public-domain network relied on by emergency managers across the United States. USGS gauges report water level (gage height) and streamflow (discharge), typically every 15 minutes.

Environment and Climate Change Canada (ECCC) GeoMet. Weather layers — radar, precipitation, and snow water equivalent — come from ECCC's GeoMet services, including the HRDPS high-resolution forecast model.

NOAA / National Weather Service. For U.S. locations, weather conditions and forecasts come from api.weather.gov.

We do not generate readings ourselves (with the exception of our own Nibi sensor hardware, which is always clearly labelled as a sensor source). If a number appears on a location page, it traces back to one of the networks above, and the source is named on the page.

What we do with the readings

Everything is stored in SI units. Feet become metres at ingest (×0.3048); cubic feet per second become cubic metres per second (×0.0283). Your display preference — feet or metres — is applied only at the last moment, on your screen. This means a Canadian gauge and a Texan gauge live in the same mathematical world and can be compared honestly.

We build daily records. Agencies publish readings every few minutes to every hour. We roll these into daily summaries per station — the daily mean, high, and low — which become the raw material for everything historical. For USGS stations this is genuinely additive work: the USGS publishes daily-mean discharge but not daily-mean gage height, so our daily level records are computed by us from the full instantaneous archive.

We compute a baseline for every calendar day. For each station, we take every July 23rd (for example) in the historical record and compute the median and percentile distribution of that day's levels. That per-day baseline — not a single annual average — is what "typical" means on Nibi Metrics. A lake is compared against what this specific date usually looks like, which is the only comparison that makes sense for water that swells every spring and recedes every summer.

Deviation is the hero number. The big number on a location page is the difference between the current reading and the median for today's date. "+2.1 ft above typical" carries meaning that a raw elevation like "223.94 m" never could. Percentile bands (normal, watch, high, low) come from the same per-day distributions.

Records are computed from complete daily records. "All-time high" on a Records & Range card is the highest completed daily record in the station's archive. During a live flood, today's reading can exceed it — the page will tell you a new record is being set, because the archive can only contain finished days.

Rate of change is measured, not modelled. The RISING / FALLING / STEADY verdict compares recent successive readings from the gauge itself. When the water is confirmed flat, we say steady — a measured zero, not an absence of data.

Forecasts

Our water-level forecast (Model v0.9) is our own statistical model. It combines the station's recent trajectory, its historical behavior for the season, and forecast precipitation from HRDPS. It is honest about itself: every forecast card carries a confidence label, and low-confidence forecasts say so. Forecasts are guidance, not guarantees — during rapidly evolving floods, always follow official emergency information.

Alerts

Weather alerts shown on Nibi Metrics are official ECCC public alerts, refreshed continuously and matched to locations geographically. We display summaries and link to the full official advisories. User-configured water-level alerts are evaluated against the same readings and baselines described above.

What we don't do

We don't smooth, adjust, or "correct" agency readings. We don't interpolate missing data and present it as measured. When a gauge goes quiet, the page says so — a stale badge appears rather than a silently frozen number. And when our own model is uncertain, the label says that too.

Questions

The data pipeline is the product, and we're glad to explain any part of it. Reach out through the contact page — if something on a location page doesn't make sense, that's a bug in our explanation even when the number is right.