Advancing river modelling in ungauged basins using satellite remote sensing: the case of the Ganges-Brahmaputra-Meghna basin

River modelling is the process of simulating the water flow dynamics of a stream network against time-varying boundary conditions. Such river models are often an important component of any flood forecasting system that forecasts river levels in flood-prone regions. However, large river basins such as the Ganges, Brahmaputra, and Meghna (GBM), Indus, Irrawaddy, Salween, Mekong, and Niger in the developing world are mostly ungauged as they lack the necessary and routine in situ measurements of river bed depth/slope, bathymetry (river cross section), flood plain mapping, and boundary condition flows for setting up of a river model. For such basins, proxy approaches relying primarily on remote-sensing data from space platforms may be the only way to overcome the lack of in situ data. In this study, we share our experience in setting up the one-dimensional River Analysis System model of the Hydrologic Engineering Center over the stream network of the GBM basin. Good-quality in situ measurements of river hydraulics (cross section, slope, flow) were available only for the basin's downstream and flood-prone region, which comprises 7% of the total basin area. For the remaining 93% of the basin area, data from the following satellite sensors were used to build a functional river model: (a) Shuttle Radar Topography Mission to derive river network and adjust river bed profiles; (b) Landsat-MODIS for updating river network and flow direction generated by elevation data; (c) radar altimetry data to build the depth versus width relationship at river locations; and (d) satellite precipitation-based hydrologic modelling of lateral flows into major rivers. We measured the success of our approach by systematically testing how well the basin-wide river model could simulate river-level dynamics at two measured downstream low-lying locations. This paper summarizes the key hurdles faced and offers a step-by-step 'rule book' approach to setting up river models for large ungauged river basins around the world. By following these rules in a systematic way, the root mean squared error for river-level simulation was reduced from 3 to 1 m. Such a guide can be useful for setting up river hydraulic models for flood forecasting systems in ungauged basins such as the Niger, Mekong, Irrawaddy, and Indus.