Sundial AI forecasts river flow across 500+ US basins
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- University of Texas at Austin conducted the study with Hydrotify LLC, and the work was published in Machine Learning: Earth (2026).
- Sundial model achieved near‑LSTM performance when forecasting river flow across a U.S. dataset of more than 500 basins.
- TSFMs (time‑series foundational models) originally trained on energy, transport and climate data were evaluated on the same river dataset.
- Snowmelt‑driven basins with strong seasonal patterns yielded the strongest AI model performance.
- Dr. Alexander Sun said the approach can provide reliable water information to regions lacking long‑term hydrological records, improving flood warnings and drought planning.
- Albert Sun contributed as an undergraduate researcher from the University of Texas at Austin.
- TSFMs capacity scales with training data size, suggesting future models that incorporate more Earth‑science records could further boost forecasting accuracy.
Why it matters: Communities in regions with sparse gauge networks gain timely flood warnings and drought forecasts, while water planners receive data‑driven insights without costly monitoring infrastructure; the study shows AI can close the information gap that has limited water‑resource decisions for decades.
