Two-Dimensional Variational Data Assimilation of SWOT derived River Discharge over Multiple Virtual Stations into the WRF-Hydro model over the Ganga River Basin, India

Distributed monitoring of river discharge remains a challenging task which require frequent measurements, particularly in data-sparse regions. In this study, we assimilated river discharge estimated using Surface Water and Ocean Topography (SWOT) observations, which were used to derive multi-station discharge distributed over a river basin. The discharge assimilation was performed using a variational method across multiple distributed virtual stations over the Ganga River and its tributaries. In the assimilation the background hydrological model estimates were optimally combined with spatio-temporally distributed SWOT-derived river discharge and associated errors. In this study, we implemented a unique dynamic window algorithm to extract water surface elevations from raw SWOT Level-2 Pixel Cloud (PIXC) data. It was designed such that data gaps and noise due to complex river reaches were excluded and data points with lower elevation uncertainty were selected. Then these elevations were used in Manning's equation with adaptive channel geometries (rectangular and parabolic) for discharge estimation over different reaches of the river. Discharge validation was carried out over multiple virtual stations across the Ganga River basin using gauge observations and GLoFAS discharge. Over stable river reaches exceptional accuracy was found (Gandhighat, Virtual Station (Ganga River): NSE>0.9, R²>0.9, RMSE