Flood predictions require several types of data:
- The amount of rainfall occurring on a real-time basis.
- The rate of change in river stage on a real-time basis, which can help indicate the severity and immediacy of the threat.
- Knowledge about the type of storm producing the moisture, such as duration, intensity and areal extent, which can be valuable for determining possible severity of the flooding.
- Knowledge about the characteristics of a river's drainage basin, such as soil-moisture conditions, ground temperature, snowpack, topography, vegetation cover, and impermeable land area, which can help to predict how extensive and damaging a flood might become.
The National Weather Service (an agency within NOAA) collects and interprets rainfall data throughout the United States and issues flood watches and warnings as appropriate. They use statistical models that incorporate USGS streamflow data to try to predict the results of expected storms. See their Advanced Hydrologic Prediction Service River Forecasts and Long-Range River Flood Risk webpages.
The USGS maintains a network of streamflow-gaging stations throughout the country.