Atmospheric rivers are some of the most challenging weather events for water managers. These powerful "rivers in the sky" can deliver months' worth of precipitation in just days, creating an acute dilemma: release water too early and risk depleting critical storage for dry months ahead; wait too long and face catastrophic flooding.
For Sonoma Water — a wholesale water supplier serving over 600,000 residents across Sonoma and Marin counties in Northern California — getting this balance right isn't just about operations. It's about protecting communities, maintaining water security, and managing the Russian River system through increasingly volatile weather patterns.
The need: Forecast accuracy in the face of uncertainty
When major atmospheric rivers come through Sonoma County, the stakes are high — but not unfamiliar. Sonoma County has experienced over $5 billion in flood damage from atmospheric rivers over the past 40 years, the highest of any county in the western U.S. Sonoma Water has deep operational expertise, built over decades of navigating both catastrophic floods and severe droughts, positioning them as leaders in the field. When an atmospheric river swept across the coast of Northern California in November 2024, Sonoma Water needed to predict inflows into key reservoirs and flood-prone areas with enough accuracy and lead time to make proactive decisions — not just react as conditions unfolded.
Extreme events can be challenging for traditional forecasting approaches. Rapidly changing storm dynamics, intensified by changing climate patterns, mean that models rooted in historical data don't always capture the full breadth of possibilities. And while meteorological forecasts can predict precipitation, translating that into actual reservoir inflows requires understanding how water moves through complex basins — especially during high-flow conditions when accuracy matters most.
HydroForecast predictions were 40% more accurate than industry standard forecasts during a real atmospheric river event.
The approach: Leading the way with FIRO and AI-driven streamflow forecasts
Since 2022, Sonoma Water has been using HydroForecast to inform operational decisions during critical events. Unlike traditional statistical models, HydroForecast uses machine learning (ML) to combine short-term weather models with long-term hydrologic signals, adapting over time by learning from past errors and seasonal variations.
During high-flow events like atmospheric rivers, HydroForecast's ensemble approach provides water managers at Sonoma Water with:
- Probabilistic inflow predictions that captured both flow magnitude and timing
- Uncertainty quantification to help operators weigh trade-offs under rapidly changing conditions
- High-resolution forecasts tailored to local basin characteristics
This is where Forecast-Informed Reservoir Operations (FIRO) becomes transformative. Rather than relying on rigid guide curves to dictate reservoir releases, FIRO integrates advanced forecasts to enable flexible, data-informed decision-making. HydroForecast fits naturally within this framework — it doesn't replace existing forecasts from NOAA or the Center for Western Weather and Water Extremes (CW3E), but supplements them with AI-powered streamflow predictions that translate atmospheric forecasts into actionable reservoir inflow data.
With HydroForecast, Sonoma Water can balance competing aspects of water management — using HydroForecast Seasonal to anticipate water supply risk, paired with HydroForecast Short-term to evaluate flood risk. The U.S. Army Corps of Engineers (USACE) is responsible for management of Lake Sonoma and Lake Mendocino when the reservoir levels get high enough to pose a flood risk to the downstream communities. Sonoma Water uses the information provided by HydroForecast Short-term and Seasonal models to determine when to hand over operations of these reservoirs to USACE. Lake Mendocino was the first officially implemented FIRO program initiated by USACE, nationwide. The results of the program speak for themselves: In a FIRO study at Lake Mendocino, the framework enabled safe storage of up to 20% more water, serving as an example for other regions facing similar highly variable water cycles.
Nathan Baskett, Hydrogeologist at Sonoma Water, explains FIRO — forecast informed reservoir operations — and how this standard is impacting water managers in the real world.
The results: Forecast insights lead to proactive operations and avoided damage
The November 2024 atmospheric river provided a compelling demonstration at Lake Mendocino of both HydroForecast's capabilities and Sonoma Water's expert decision-making under extreme conditions. This was no ordinary storm — 13 inches of rain fell in just 6 days, creating what meteorologists classified as a 1,000-year event in Santa Rosa and causing over $2 million in damages across the region.
HydroForecast timeline
- Three days before the first peak, HydroForecast was already showing strong signals when 8-12 inches of precipitation were forecasted.
- One day ahead of the first peak, HydroForecast accurately predicted a second burst of precipitation and showed closer alignment to eventual observations than River Forecast Center (RFC) forecasts.
- As the event unfolded, HydroForecast maintained close alignment with observed flows throughout multiple peaks.
Even as conditions evolved rapidly — with additional precipitation bringing a second peak — HydroForecast maintained its accuracy advantage. The model captured not just the magnitude of the peaks, but also the timing and shape of the hydrograph, including the receding limb as flows declined.
Key performance metrics
- HydroForecast predicted the peak flow 24 hours in advance, giving operators critical lead time to prepare
- 40% more accurate than the RFC predictions during this high-flow event
Armed with these accurate forecasts, Sonoma Water's team took decisive action to protect the community. They coordinated the transfer of reservoir operations to USACE, activated the Flood Forecast Hotline to keep residents informed, and removed critical equipment out of harm's way — proactive measures that helped minimize over $2 million in regional damages. Their quick decision-making exemplifies how technical tools and operational expertise must work together during high-stakes events.
The big picture: Building resilience through better forecasts
Atmospheric rivers are responsible for 30-50% of annual precipitation in parts of the western U.S., but they're becoming harder to predict using historical patterns alone. The U.S. Bureau of Reclamation recently initiated the integration of HydroForecast at multiple reservoirs, and Sonoma Water's successful navigation of the November 2024 event demonstrates that advanced forecasting tools paired with experienced operators can transform how agencies manage increasingly unpredictable weather patterns. By integrating AI-driven forecasts with proven FIRO frameworks, agencies are showing what's possible: protecting communities from floods while securing water supply for dry periods.
Learn more about managing atmospheric rivers
Interested in how forecast-informed reservoir operations can transform your water management approach?
