Comment on hess-2022-111

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.