A recent study reveals that when a specialized AI is combined with the National Water Model (a dataset produced by NOAA), the hybrid system becomes four to six times more accurate in forecasting streamflow. The tool significantly boosts AI flood prediction accuracy, offering more reliable forecasts of where floods will occur.
The AI is named Errorcastnet and it is able to find and correct errors within the national model. By training on historical rainfall and flood data, it learned where the model typically fails. The researchers who completed the study emphasize that while the AI is an effective tool, it cannot replace the decades of research and physics-based models that account for key variables like elevation, vegetation, and drainage.
Bettering NOAA's forecasting model could reduce the potential economic impacts of floods by allowing businesses and communities to better prepare.
Read the full study to learn more about the methodology and findings.
