From weather data to river runoff: using spatiotemporal convolutional networks for discharge forecasting

Börgel, Florian; Karsten, Sven; Rummel, Karoline; Gräwe, Ulf

doi:https://doi.org/10.5194/gmd-18-2005-2025

Articles | Volume 18, issue 6

https://doi.org/10.5194/gmd-18-2005-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-18-2005-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 18, issue 6

Model experiment description paper

|

27 Mar 2025

Model experiment description paper |

| 27 Mar 2025

From weather data to river runoff: using spatiotemporal convolutional networks for discharge forecasting

Florian Börgel, Sven Karsten, Karoline Rummel, and Ulf Gräwe

Data sets

Supporting data ocean model GMD submission: From Weather Data to River Runoff: Leveraging Spatiotemporal Convolutional Networks for Comprehensive Discharge Forecasting F. Börgel https://doi.org/10.5281/zenodo.13365099

Model code and software

florianboergel/runoff_prediction: gmd (Version v1) F. Börgel https://doi.org/10.5281/zenodo.13910136

Elements of the Modular Ocean Model (MOM) (2012 release with updates) GFDL Ocean Group Technical Report No. 7 (https://github.com/mom-ocean/MOM5) S. Griffies https://mom-ocean.github.io/assets/pdfs/MOM5_manual.pdf

Short summary

Forecasting river runoff, which is crucial for managing water resources and understanding climate impacts, can be challenging. This study introduces a new method using convolutional long short-term memory (ConvLSTM) networks, a machine learning model that processes spatial and temporal data. Focusing on the Baltic Sea region, our model uses weather data as input to predict daily river runoff for 97 rivers.