Forecasting through extreme weather

Atmospheric rivers, hurricanes, and the floods they unleash are getting more intense — and harder to forecast. HydroForecast gives operators advantages when it matters most including the lead time to act before the peak, the accuracy to trust the call, and forecasts that keep running when gauges fail.

Storm with palm trees bending in wind alongside a HydroForecast peak flow prediction chart

Act before the peak

The difference between a managed drawdown and an emergency is lead time. HydroForecast detects the potential for extreme flows days earlier than traditional forecasts — across atmospheric rivers, hurricanes, and rapid snowmelt — so operators can stage crews, coordinate downstream, and make reservoir calls before the water arrives.

10 days

advance warning during Dec 2025 PNW sequence

See the full results →

7 days

ahead of Helene's peak and +4 days over NOAA RFC

See the full results →
HydroForecast chart showing early warning 5 days before Hurricane Helene's peak, compared to a flat NOAA RFC forecast
HydroForecast vs. NOAA RFC — forecast issued 5 days before Helene's peak. HydroForecast predicted the surge; the RFC showed near-zero flow.

No gauge, no problem

Extreme weather knocks out the very infrastructure forecasts depend on — gauges fail, telemetry drops, and the rivers most at risk often have no monitoring at all. HydroForecast keeps running through the outage and reaches ungauged basins, so dam-safety decisions never go dark at the worst possible moment.

When Hurricane Helene took the USGS gauge at the Nolichucky River offline, HydroForecast kept running without interruption — predicting flows within 5% of the observed peak at the exact moment operators needed it most.

Map of USGS streamflow monitoring locations across the United States, showing gaps in coverage
USGS streamflow gauge locations — white circles mark "current streamflow unavailable." Many rivers at highest flood risk have no coverage at all.

HydroForecast gives us crucial insights into upcoming storm and heavy rain events. The dashboard is an integral part of our emergency response workflows and supports our commitment to public safety — we also use it to plan hydrology operational work and look forward to applying insights to inform drought response.

Auckland Council Kolt Johnson, Team Manager, Hydrology and Environmental Data

Built for the unprecedented

Extreme events generate flows with no historical precedent — exactly where statistical models break down. HydroForecast uses a unique AI modeling approach that's built to stay accurate when rainfall lands on saturated soils and rivers run higher than the record books have ever seen.

Make data-driven decisions when every minute counts

During a demanding atmospheric river season, HydroForecast outperformed the RFC by 27% in mean absolute error — roughly 15,000 acre-feet of water. The advantage held through the hardest events in the sequence, exactly when accurate forecasts matter most.

27%

more accurate than RFC

See the full results →

Stay accurate when there's no historical precedent to fall back on

In the 1–5 day window before Helene's unprecedented peak, HydroForecast achieved 33% lower error than traditional systems — accuracy that holds precisely when the stakes are highest and historical precedent runs out.

33%

lower error

See the full results →

Go deeper: Hurricanes, flooding, and more


Ready to see HydroForecast in action?

Talk to our team