Preprints
https://doi.org/10.5194/hess-2024-183
https://doi.org/10.5194/hess-2024-183
18 Jul 2024
 | 18 Jul 2024
Status: a revised version of this preprint is currently under review for the journal HESS.

Neural networks in catchment hydrology: A comparative study of different algorithms in an ensemble of ungauged basins in Germany

Max Weißenborn, Lutz Breuer, and Tobias Houska

Abstract. This study presents a comparative analysis of different neural network models, including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) in predicting discharge within ungauged basins in Hesse, Germany. All models were trained on 54 catchments with 28 years of daily meteorological data, either including or excluding 11 static catchment attributes. The training process of each model scenario combination was repeated 100 times, using a Latin Hyper Cube Sampler for the purpose of hyperparameter optimisation with batch sizes of 256 and 2048. The evaluation was carried out using data from 35 additional catchments (6 years) to ensure predictions in basins that were not part of the training data. This evaluation assesses predictive accuracy, computational efficiency concerning varying batch sizes and input configurations and conducted a sensitivity analysis of various hydrological and meteorological. The findings indicate that all examined artificial neural networks demonstrate significant predictive capabilities, with a CNN model exhibiting slightly superior performance, closely followed by LSTM and GRU models. The integration of static features was found to improve performance across all models, highlighting the importance of feature selection. Furthermore, models utilising larger batch sizes displayed reduced performance. The analysis of computational efficiency revealed that a GRU model is 41 % faster than the CNN and 59 % faster than the LSTM model. Despite a modest disparity in performance among the models (<3.9 %), the GRU model's advantageous computational speed renders it an optimal compromise between predictive accuracy and computational demand.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.
Max Weißenborn, Lutz Breuer, and Tobias Houska

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • CC1: 'Comment on hess-2024-183', John Ding, 23 Jul 2024
    • CC2: 'Reply on CC1', Max Weißenborn, 26 Jul 2024
      • CC3: 'Reply on CC2', John Ding, 16 Aug 2024
        • CC4: 'Reply on CC3', John Ding, 16 Aug 2024
  • RC1: 'Comment on hess-2024-183', Anonymous Referee #1, 05 Aug 2024
    • AC1: 'Reply on RC1', Max Weißenborn, 10 Oct 2024
    • AC4: 'Reply on RC1', Max Weißenborn, 10 Oct 2024
  • RC2: 'Comment on hess-2024-183', Anonymous Referee #2, 15 Aug 2024
    • AC2: 'Reply on RC2', Max Weißenborn, 10 Oct 2024
    • AC3: 'Reply on RC2', Max Weißenborn, 10 Oct 2024
Max Weißenborn, Lutz Breuer, and Tobias Houska
Max Weißenborn, Lutz Breuer, and Tobias Houska

Viewed

Total article views: 932 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
647 206 79 932 17 11
  • HTML: 647
  • PDF: 206
  • XML: 79
  • Total: 932
  • BibTeX: 17
  • EndNote: 11
Views and downloads (calculated since 18 Jul 2024)
Cumulative views and downloads (calculated since 18 Jul 2024)

Viewed (geographical distribution)

Total article views: 886 (including HTML, PDF, and XML) Thereof 886 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 05 Dec 2024
Download
Short summary
Our study compares neural network models for predicting discharge in ungauged basins. We evaluated Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) using 28 years of weather data. CNN showed the best accuracy, while GRU were faster and nearly as accurate. Adding static features improved all models. The research enhances flood forecasting and water management in regions lacking direct measurements, offering efficient and accurate predictive tools.