High-Resolution Global Streamflow Dataset from 1980 - 2020 for 2.94 Million Rivers Using the Physics-Embedded δHBV2–δMC2 Model

This dataset provides global-scale streamflow simulations generated with the physics-embedded, high-resolution δHBV2–δMC2 model, developed by the Multi-scale Hydrology, Processes, and Intelligence (MHPI) team at The Pennsylvania State University, led by Dr. Chaopeng Shen’s Hydrologic Deep Learning and Modeling group. This data is provided as is, and we do not assume any responsibility as a result of the use of this data.

This dataset is a direct result of Ji et al., 2025 described below, which was trained from 1980-2000 and simulated from 1980-2020, and built upon the work in Song et al., 2025. Due to Zenodo's storage limitations, a complete copy of the global simulation has been uploaded to hydroshare (http://www.hydroshare.org/resource/6c8191d3613c4477b717be41c81a4372), and this zenodo repo contains files only for USA and Europe.

Please cite these two papers if you find the data to be of use (* indicates MHPI group members, italic indicates corresponding author) and refer to the Zenodo DOI rather than the HydroShare link when acknowledging the data source:

Ji, Haoyu*, Yalan Song*, Tadd Bindas*, Chaopeng Shen*, Yuan Yang, Ming Pan, Jiangtao Liu*, Farshid Rahmani*, Ather Abbas, Hylke Beck, Kathryn Lawson* and Yoshihide Wada (2025). Distinct hydrologic response patterns and trends worldwide revealed by physics-embedded learning. Nature Communications, doi.org/10.1038/s41467-025-64367-1

Song, Yalan*, Tadd Bindas*, Chaopeng Shen*, Haoyu Ji*, Wouter J. M. Knoben, Leo Lonzarich*, Martyn P. Clark, Jiangtao Liu*, Katie van Werkhoven, Sam Lemont, Matthew Denno, Ming Pan, Yuan Yang, Jeremy Rapp, Mukesh Kumar, Farshid Rahmani*, Cyril Thébault, Richard Adkins, James Halgren, Trupesh Patel, Arpita Patel, Kamlesh Sawadekar*, and Kathryn Lawson* (2025). High-resolution national-scale water modeling is enhanced by multiscale differentiable physics-informed machine learning. Water Resources Research, doi: 10.1029/2024WR038928

More information on our research group and publications can be found at www.mhpi.info, and our publicly-available codes are also available on Github at www.github.com/mhpi. A wiki with summaries, benchmarks, and more information is available at https://mhpi.github.io.

To facilitate data access, the global river network has been partitioned into zones based on the MERIT Flowlines framework. Users can download subsets corresponding to their regions of interest. The MERIT basin delineations are available at: https://www.reachhydro.org/home/params/merit-basins.

Note: The model simulates other variables like ET, soil moisture and snow water equivalent, etc., but are not evaluated. If you are interested, please let us know.

See the README file for more information.