The rainfall-to-runoff problem
River flooding begins with rainfall, but the relationship between precipitation and river level is complex. The same amount of rain falling on saturated soil produces far more runoff than on dry soil. Urban watersheds with impervious surfaces respond faster than forested watersheds. Snowmelt can add massive volumes of water to rivers with no current rainfall. Hydrological forecasting is the science of modeling these complex watershed responses.
The core challenge is converting a spatial field of rainfall (or snowmelt) into a predicted hydrograph — the time series of water level or discharge at a specific point on a river. This requires modeling how water moves across the land surface, infiltrates into soil, moves through subsurface pathways, and eventually concentrates in stream channels.
How hydrological models work
Modern hydrological models divide a watershed into spatial units (sub-basins or grid cells) and simulate the water balance in each unit. Rainfall enters the system, some infiltrates into soil, some runs off across the surface, some evaporates. The infiltrated water moves through soil layers and eventually reaches streams through subsurface flow. Surface runoff is routed through the channel network to the forecast point.
The National Weather Service operates the National Water Model, a continental-scale hydrological model that simulates streamflow for 2.7 million river reaches across the entire United States. The model runs every hour, ingesting the latest radar rainfall estimates, weather forecasts, and soil moisture observations to produce river stage forecasts at thousands of locations.
Forecast lead times and accuracy
River flood forecast lead time depends on watershed size. Small urban streams may respond to rainfall in 1–3 hours, giving minimal forecast lead time. Medium-sized rivers in hilly terrain provide 6–24 hours. Large river systems like the Mississippi, Missouri, or Ohio may take days to weeks to respond to upstream rainfall, providing multi-day forecast lead times.
Forecast accuracy depends on the quality of precipitation input (observed and forecast), the model's calibration to the specific watershed, and the initial soil moisture conditions. For well-instrumented watersheds with calibrated models, river stage forecasts are typically accurate to within 0.5–1 meter for lead times of 1–3 days. Beyond 3 days, accuracy degrades because it depends on weather forecast accuracy.
The role of upstream gauge data
For large rivers, the most powerful forecast input is the observed water level at upstream gauge stations. If a river is cresting at an upstream location, the downstream forecast can use that observed crest and river routing models to predict when and at what level the crest will arrive downstream. This produces highly accurate forecasts with lead times determined by the travel time between gauges.
The USGS stream gauge network is the foundation of this capability. During major flood events, gauge data is updated every 15 minutes and made available through the USGS WaterWatch and NWS Advanced Hydrologic Prediction Service (AHPS) websites. PlanetSentry and other monitoring platforms can reference these forecasts to provide context for flood events displayed on the globe.
Ensemble forecasting and uncertainty
Because precipitation forecasts are uncertain — especially beyond 2–3 days — hydrological forecast systems increasingly use ensemble approaches. Instead of running one model with one rainfall forecast, they run the model with 20–50 different plausible rainfall scenarios derived from ensemble weather forecasts. The result is a range of possible river level outcomes with associated probabilities.
This probabilistic approach is more useful for decision-making than a single deterministic forecast. An emergency manager can see that there is a 30% chance of major flooding and a 70% chance of moderate flooding, rather than receiving a single river level estimate that may not capture the uncertainty. The NWS Hydrologic Ensemble Forecast Service (HEFS) produces these probabilistic forecasts operationally.
Global flood forecasting systems
The Global Flood Monitoring System (GFMS) and the European Global Flood Awareness System (GloFAS) provide flood forecasts at continental to global scales. These systems use satellite precipitation data, global weather model forecasts, and hydrological models to estimate flood conditions worldwide, including in countries without dense ground monitoring networks.
GDACS uses these global systems to assess flood severity and issue alerts. When PlanetSentry displays a GDACS flood alert, the underlying assessment draws on global flood modeling that estimates the extent, depth, and population exposure of the flood event — providing actionable intelligence even for floods in data-sparse regions of the world.