Three forecast horizons designed for organizations' decision-making:
HydroForecast swept a year long short-term flow forecasting competition hosted by the United States Bureau of Reclamation, CEATI and hydropower utilities and verified by RTI International.
In every competition region, HydroForecast was more accurate further into the future and provided better insight in both drought conditions and 1000-year storms.
HydroForecast learns to represent hydrologic processes by identifying relationships between satellite observations, basin characteristics, meteorological forecasts, and streamflow measurements.
Our approach is distinct from purely data-driven machine learning models in that we use physical science to guide and constrain the relationships between inputs and streamflow predictions. We believe that the best model is one grounded in scientific principles and enhanced by finding deep connections in the relationships between data.
HydroForecast’s Core Model has been trained on a wide range of hydrologic conditions and landscapes, and can be tuned for specific locations anywhere on Earth.