California's stream gauge network is foundational to modern water management, but gauges are expensive to install and maintain, geographically biased toward large accessible rivers, and vulnerable to physical damage and telemetry failure at the worst possible moments. Significant monitoring gaps remain across headwaters, smaller tributaries, and remote catchments that physical infrastructure alone cannot fill at any practical scale or cost.
Machine learning is changing what's operationally possible. Models trained across hundreds of hydrologically diverse basins can now generate reliable streamflow forecasts at ungauged locations without local calibration data, extending flow information to reaches that may never see a physical gauge. In one recent application of this technology, Upstream Tech generated a daily historical streamflow dataset for every river reach across California, demonstrating the breadth and utility of this approach at a statewide scale.
HydroForecast, built by Upstream Tech, is an AI-powered probabilistic streamflow forecasting tool used by hydropower producers, utilities, and government agencies to support more informed water management decisions. Backed by experts in hydrology and machine learning, HydroForecast extends actionable forecasting even to ungauged locations, helping fill critical data gaps for water professionals across California and beyond.
In this session, Alex Truby will walk through how ungauged forecasting works and where it is being applied in California water management, including:
- Forecast-Informed Reservoir Operations (FIRO) and FIRO-MAR: Reservoirs operating under FIRO frameworks depend on accurate inflow forecasts to inform pre-storm release decisions. Ungauged forecasting extends that capability to tributary catchments where gauges don't exist and supports managed aquifer recharge (MAR) programs that require flow estimates across unmonitored reaches to identify and time recharge opportunities.
- Water accounting and allocation: Generating consistent flow records across unmonitored reaches supports water rights administration and environmental flow compliance in basins with historically sparse gauge coverage.
- Telemetry QA/QC: Modeled flows can serve as a physically grounded reference to flag suspect gauge readings and fill data gaps during sensor outages, precisely when accurate data matters most.
Why this matters
Data gaps don't have to be decision gaps. For water managers, regulators, and environmental practitioners working across California's diverse and often remote river network, the ability to access reliable flow information beyond where gauges happen to exist has meaningful implications for environmental flow compliance, flood preparedness, water allocation, and long-term planning. If your work depends on understanding streamflow conditions across California watersheds, this session is for you.
