Preprints
https://doi.org/10.5194/hess-2022-111
https://doi.org/10.5194/hess-2022-111
 
25 Apr 2022
25 Apr 2022
Status: this preprint is currently under review for the journal HESS.

A graph neural network approach to basin-scale river network learning: The role of physics-based connectivity and data fusion

Alexander Y. Sun1, Peishi Jiang2, Zong-Liang Yang3, Yangxinyu Xie4, and Xingyuan Chen2 Alexander Y. Sun et al.
  • 1Bureau of Economic Geology, The University of Texas at Austin, Austin, TX, USA
  • 2Pacific Northwest National Laboratory, Richland, WA, USA
  • 3Department of Geological Sciences, The University of Texas at Austin, Austin, TX, USA
  • 4Department of Computer Science, The University of Texas at Austin, Austin, TX, USA

Abstract. Rivers and river habitats around the world are under sustained pressure from anthropogenic activities and the changing global environment. Our ability to quantify and manage the river states in a timely manner is critical for protecting the public safety and natural resources. Vector-based river network models have enabled modeling of large river basins at increasingly fine resolutions, but are computationally demanding. This work presents a multistage, physics-guided, graph neural network (GNNs) approach for basin-scale river network learning and stream forecasting. GNN models are pretrained using a high-resolution vector-based river network model, and then fine-tuned with in situ streamflow observations, after which a post-processing data fusion step is proposed to propagate residuals over the entire network to correct predictions. The GNN-based framework is demonstrated over a snow-dominated watershed in the western U.S. consisting of 552 reaches. A series of experiments are performed to test different training and imputation strategies. Results show the trained GNN model can effectively serve as a surrogate model of the process-based model with high accuracies, with the median Kling–Gupta efficiency (KGE) greater than 0.97. Application of the graph-based data fusion further reduces mismatch between the GNN model and observations, with as much as 50 percent KGE improvement over cross-validation gages. Additionally we exploit and demonstrate a graph coarsening procedure that achieves comparable predicting skills at only a fraction of training cost, thus providing important insights on the degree of physical realism needed for developing large-scale GNN-based river network models.

Alexander Y. Sun et al.

Status: open (until 01 Jul 2022)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse

Alexander Y. Sun et al.

Alexander Y. Sun et al.

Viewed

Total article views: 391 (including HTML, PDF, and XML)
HTML PDF XML Total Supplement BibTeX EndNote
252 134 5 391 18 3 2
  • HTML: 252
  • PDF: 134
  • XML: 5
  • Total: 391
  • Supplement: 18
  • BibTeX: 3
  • EndNote: 2
Views and downloads (calculated since 25 Apr 2022)
Cumulative views and downloads (calculated since 25 Apr 2022)

Viewed (geographical distribution)

Total article views: 364 (including HTML, PDF, and XML) Thereof 364 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 27 May 2022
Download
Short summary
High-resolution river modeling is of great interest to local governments and stakeholders for flood hazard mitigation. This work presents a physics-guided, machine learning (ML) framework for combining the strengths of high-resolution process-based river network models with a graph-based ML model capable of modeling spatiotemporal processes. Results show the ML model can approximate the dynamics of the process model with high fidelity, and data fusion further improves the forecasting skill.