Assessing Spatial Redistribution of River Runoff Using a Relative Centroid Index (RCI): Insights from Multi-Model and Remote Sensing Data Comparisons

Anthropogenic activities, such as water withdrawals and inter-basin transfers, cause significant spatial redistribution of river runoff along channel networks. However, this process is poorly constrained in large-scale hydrological models (LHMs) traditionally calibrated against streamflow time series only at basin outlets. To address this gap, we employ a novel Relative Centroid Index (RCI), a normalized metric quantifying the upstream/downstream shift of the runoff "center of mass" within a basin, which serves as a novel metric to evaluate how well models capture such spatial footprints. We first calculate benchmark RCI values (RCI_gauge) at over 200 globally distributed basins with sufficient gauge density. We then evaluate the capability of four major model-based global river discharge products to replicate these observed RCI patterns at gauge locations. They involve GRADES, its enhanced versions GRFR and GRADES-hydroDL, and GRDR which adds a river width data assimilation module. Furthermore, we explore the potential of remote sensing (Landsat, SWOT) to provide complementary spatial distribution information, despite potential biases in absolute magnitude. Preliminary analysis suggests systematic biases in model-simulated RCI particularly in highly human regulated basins, while remote sensing shows promise in capturing relative spatial patterns. This work provides a new framework to diagnose spatial inaccuracies in LHMs and highlights the value of multi-source observations for improving the representation of human-altered hydrological processes.