CAMELS-G Graph Dataset

CAMELS-G is a large-scale graph dataset comprising more than 500 U.S. basin networks, nearly 100,000 river nodes, and approximately six billion data points. It was generated using public-domain hydrofabrics datasets to delineate river networks and the US National Water Model (NWM) dataset to provide sub-daily streamflow training data. A detailed study using the CAMELS-G is described in "Learning Geometry-Aware Streamflow Forecast Models from a Large-Scale River Graph Dataset" Journal of Hydrology, 2026, 136059, ISSN 0022-1694, https://doi.org/10.1016/j.jhydrol.2026.136059.10.1016/j.jhydrol.2026.136059