Articles | Volume 26, issue 13
https://doi.org/10.5194/hess-26-3377-2022
https://doi.org/10.5194/hess-26-3377-2022
Research article
 | 
05 Jul 2022
Research article |  | 05 Jul 2022

Deep learning rainfall–runoff predictions of extreme events

Jonathan M. Frame, Frederik Kratzert, Daniel Klotz, Martin Gauch, Guy Shalev, Oren Gilon, Logan M. Qualls, Hoshin V. Gupta, and Grey S. Nearing

Data sets

CAMELS return period analysis Jonathan M. Frame https://doi.org/10.4211/hs.c7739f47e2ca4a92989ec34b7a2e78dd

MC-LSTM, model runs Jonathan M. Frame https://doi.org/10.4211/hs.d750278db868447dbd252a8c5431affd

The CAMELS data set: catchment attributes and meteorology for large-sample studies N. Addor, A. J. Newman, N. Mizukami, and M. P. Clark https://doi.org/10.5065/D6G73C3Q

NOAA National Water Model CONUS Retrospective Dataset NOAA National Water Model CONUS Retrospective Dataset https://registry.opendata.aws/nwm-archive

Model code and software

Code for calibrating SAC-SMA Grey S. Nearing https://github.com/Upstream-Tech/SACSMA-SNOW17

NeuralHydrology (https://github.com/neuralhydrology/neuralhydrology) F. Kratzert, M. Gauch, G. Nearing,, and D.Klotz https://doi.org/10.21105/joss.04050

SPOTting Model Parameters Using a Ready-Made Python Package T. Houska, P. Kraft, A. Chamorro-Chavez, and L.Breuer https://pypi.org/project/spotpy/

jmframe/mclstm_2021_extrapolate: Submit to HESS 5_August_2021 Jonathan Frame https://doi.org/10.5281/zenodo.5165216

Log-Pearson Flood Flow Frequency using USGS 17B J. Burkey https://www.mathworks.com/matlabcentral/fileexchange/22628-log-pearson-flood-flow-frequency-using-usgs-17b

Download

The requested paper has a corresponding corrigendum published. Please read the corrigendum first before downloading the article.

Short summary
The most accurate rainfall–runoff predictions are currently based on deep learning. There is a concern among hydrologists that deep learning models may not be reliable in extrapolation or for predicting extreme events. This study tests that hypothesis. The deep learning models remained relatively accurate in predicting extreme events compared with traditional models, even when extreme events were not included in the training set.