Hydrology forecast & decision support, for a less predictable climate.
Streamflow predicted on a 72 hour horizon, then turned into a decision. Rainfall ingested from NOAA, ECMWF, and radar, reconciled with your ground gauges. Confidence bands shown by default.
The baseline most hydrology models were calibrated against is no longer the world you operate in. Storms arrive sooner, drought tails run longer, and the historical record is a guide rather than a rule. Forecasting at utility grade now means accepting that uncertainty, quantifying it, and turning it into a decision — not hiding it.
A 30 year baseline is one season behind the weather.
A reservoir release schedule built on a 30 year baseline is operating one season behind the weather. Excel based release tables cannot ingest the data fast enough. National weather products do not arrive in a format the control room can act on. The result is release errors, conservative releases, and end of season shortfalls that nobody wanted.
What the system does.
Multi source weather assimilation
NOAA (HRRR, NEXRAD), ECMWF, and GFS reconciled with your in basin gauges. Disagreement between products is shown, not averaged away.
72 hour streamflow forecast
LSTM and Transformer models forecast inflow at the reservoir, the diversion, and the gauging station. Confidence bands at 50%, 80%, and 95% are returned by default.
Decision support, not just a number
Release schedules update with the forecast. The Compliance Agent flags downstream environmental flow obligations and the model recommends a gate schedule. The operator stays in control.
What-if & historical replay
Run a scenario against your Digital Twin, or replay any past storm against the current model — with citations to the actual data each answer relied on.
From rainfall to a gate schedule.
- 01
Ingest
Pulls NOAA and HRRR rainfall forecasts alongside your in-basin gauges and local sensor data.
- 02
Visualise
The hydrologist sees a watershed dashboard — a heat map of projected catchment saturation, 72 hours out.
- 03
Simulate
Run a what-if against your Digital Twin: “if streamflow reaches 500 cfs, what happens to Reservoir X?”
- 04
Act
The system recommends a gate-opening schedule that balances downstream safety against storage efficiency.
How it deploys.
Read only integration with your gauges, historians, and national weather feeds. Sovereign deployment in customer VPC or on premise. AES 256, RBAC, SSO and SAML. 30 day pilot, scoped to one basin, with a documented outcome at the end.
Schedules that keep up with the weather.
Fewer release errors and fewer conservative over-releases, with downstream environmental-flow obligations met. We’re early — we’ll publish verified results as our pilots produce them.
Common questions.
Where does the weather data come from?
+NOAA products (HRRR, NEXRAD radar), ECMWF, and GFS, reconciled with your in-basin gauges. Where the products disagree, the disagreement is shown rather than averaged away.
How accurate is the 72 hour forecast?
+Forecasts are returned with confidence bands at 50%, 80%, and 95%. The model publishes its own skill score per basin rather than a single headline number, so you can see where it is and is not reliable.
Does it replace our release scheduling?
+No. Release schedules update with the forecast and the Compliance Agent flags downstream environmental-flow obligations, but the operator stays in control of every release.
How long does a pilot take?
+30 days on one basin. You can also replay a past season to see how the model would have behaved against data you already have.
Run your last difficult season through the model.
Compare what the forecast would have said with what the operator actually saw.