Global Daily Discharge Estimation Based on Grid Long Short-Term Memory (LSTM) Model and River Routing

Corresponding peer-reviewed publication

• Yang, Y., Feng, D., Beck, H.E., Hu, W., Ather, A., Sengupta, A., Delle Monache, L., Hartman, R., Lin, P., Shen, C. and Pan, M., 2025. Global Daily Discharge Estimation Based on Grid Long Short-Term Memory (LSTM) Model and River Routing. Water Resources Research, 61, e2024WR039764. https://doi.org/10.1029/2024WR039764 
• For any updates, please refer to the GRADES-hydroDL website: https://www.reachhydro.org/home/records/grades-hydrodl.

When using any of the files in this dataset, please cite both the article as mentioned above and the dataset herein. 

Summary

This dataset contains input files for developing the GRADES-hydroDL (global reach level daily discharge based on machine learning and river routing model) dataset, evaluation scripts, and final evaluation metrics.

• GRADES-hydroDL.pdf: The details of the GRADES-hydroDL dataset, including overview, download links, instructions, etc.
• input.zip:  Input files for LSTM training and application, including information and attributes of selected basins for LSTM training, 10-fold cross-validation gauges, and basic information of the global 0.25-degree grids used for LSTM application.
• metrics.zip: All evaluation results of all experiments used in the article.
• simulation.zip: Daily simulation of training gauges. Global simulations (GRADES-hydroDL), please see GRADES-hydroDL.pdf.
• evaluation_script:  Scripts for calculating metrics and plotting figures.
• UCSD_LICENSE.md