Articles | Volume 23, issue 12
Hydrol. Earth Syst. Sci., 23, 5089–5110, 2019
https://doi.org/10.5194/hess-23-5089-2019
Hydrol. Earth Syst. Sci., 23, 5089–5110, 2019
https://doi.org/10.5194/hess-23-5089-2019

Research article 17 Dec 2019

Research article | 17 Dec 2019

Towards learning universal, regional, and local hydrological behaviors via machine learning applied to large-sample datasets

Frederik Kratzert et al.

Data sets

Pre-trained models F. Kratzert https://doi.org/10.4211/hs.83ea5312635e44dc824eeb99eda12f06

Benchmark models F. Kratzert https://doi.org/10.4211/hs.474ecc37e7db45baa425cdb4fc1b61e1

CAMELS extended Maurer forcings F. Kratzert https://doi.org/10.4211/hs.17c896843cf940339c3c3496d0c1c077

Model code and software

Code to reproduce paper experiments/results F. Kratzert https://doi.org/10.5281/zenodo.3530884

Download
Short summary
A new approach for regional rainfall–runoff modeling using long short-term memory (LSTM)-based models is presented and benchmarked against a range of well-known hydrological models. The approach significantly outperforms regionally calibrated hydrological models but also basin-wise calibrated models. Furthermore, we propose an adaption of the LSTM that allows us to extract the learned catchment understanding of the model and show that it matches our hydrology expert knowledge.