11 Aug 2021

11 Aug 2021

Review status: this preprint is currently under review for the journal HESS.

Karst spring discharge modeling based on deep learning using spatially distributed input data

Andreas Wunsch1, Tanja Liesch1, Guillaume Cinkus2, Nataša Ravbar3, Zhao Chen4, Naomi Mazzilli5, Hervé Jourde2, and Nico Goldscheider1 Andreas Wunsch et al.
  • 1Karlsruhe Institute of Technology (KIT), Institute of Applied Geosciences, Hydrogeology, Kaiserstr. 12, 76131 Karlsruhe, Germany
  • 2HydroSciences Montpellier (HSM), Université de Montpellier, CNRS, IRD, 34090 Montpellier, France
  • 3ZRC SAZU, Karst Research Institute, Titov trg 2, 6230 Postojna, Slovenia
  • 4Environmental Resources Management, Siemensstr. 9, 63263 Neu-Isenburg, Germany
  • 5UMR 1114 EMMAH (AU-INRAE), Université d’Avignon, 84000 Avignon, France

Abstract. Despite many existing approaches, modeling karst water resources remains challenging and often requires solid system knowledge. Artificial Neural Network approaches offer a convenient solution by establishing a simple input-output relationship on their own. However, in this context, temporal and especially spatial data availability is often an important constraint, as usually no or few climate stations within a karst spring catchment are available. Hence spatial coverage is often unsatisfying and can introduce severe uncertainties. To avoid these problems, we use 2D-Convolutional Neural Networks (CNN) to directly process gridded meteorological data followed by a 1D-CNN to perform karst spring discharge simulation. We investigate three karst spring catchments in the Alpine and Mediterranean region with different meteorologic-hydrological characteristics and hydrodynamic system properties. We compare our 2D-models both to existing modeling studies in these regions and to 1D-models, which use climate station data, as it is common practice. Our results show that our models are excellently suited to model karst spring discharge and rival the simulation results of existing approaches in the respective areas. The 2D-models learn relevant parts of the input data and by performing a spatial input sensitivity analysis we can further show their potential for karst catchment localization and delineation.

Andreas Wunsch et al.

Status: open (until 27 Oct 2021)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on hess-2021-403', Kuo-Chin Hsu, 03 Sep 2021 reply
  • RC2: 'Comment on hess-2021-403', Anonymous Referee #2, 20 Oct 2021 reply

Andreas Wunsch et al.

Model code and software

CNN_KarstSpringModeling Andreas Wunsch

Andreas Wunsch et al.


Total article views: 483 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
359 120 4 483 3 1
  • HTML: 359
  • PDF: 120
  • XML: 4
  • Total: 483
  • BibTeX: 3
  • EndNote: 1
Views and downloads (calculated since 11 Aug 2021)
Cumulative views and downloads (calculated since 11 Aug 2021)

Viewed (geographical distribution)

Total article views: 417 (including HTML, PDF, and XML) Thereof 417 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 24 Oct 2021
Short summary
Modeling complex karst water resources is difficult enough, but often there are no or too few climate stations available to deliver input data. We apply common image recognition algorithms to weather videos to simulate karst spring discharge. For this we use raster data, which are usually more readily available than weather station data. Our models can also learn the catchment area of a spring independently and may replace conventional field methods for catchment delineation in the future.