Understanding the Geophysical Sources of Uncertainty for Satellite Interferometric (SRTM)-Based Discharge Estimation in River Deltas: The Case for Bangladesh

Like most river deltas, Bangladesh represents a geographically small region with numerous crisscrossing rivers. The total number of rivers in Bangladesh exceeds 300, of which 57 rivers are transboundary. Given the widespread unavailability of flow data across the entire river basins of Ganges, Brahmaputra, and Meghna, combined with a declining measurement network and political challenges of sharing the data, satellite remote sensing of discharge has recently become a viable alternative. This study was motivated by the need to understand the geophysical sources of uncertainty of satellite interferometric-based discharge estimation in Bangladesh. A consequential goal of this study was to contextualize the understanding as a function of river´s geophysical characteristics (river width, reach averaging length, and bed/water slope) and also to explore a pragmatic approach to uncertainty reduction using water level climatology. Discharge was estimated according to the slope-area (Manning´s) method using elevation data from Shuttle Radar Topography Mission (SRTM). A high-resolution hydrodynamic (HD) model was accurately calibrated to simulate water level and flow dynamics along the river reaches of the river network and serve as reference for comparison with satellite-based estimates. It was found that satellite interferometric (SRTM)-based discharge estimates yielded estimation error variance an order smaller than the natural flow variability only if the river width was at least three times larger the width of the native resolution of satellite elevation data. Rivers narrower than this width (for SRTM, this cutoff is 270 m) yielded a coefficient of variation larger than 1 due to contamination of land elevation data in hydraulic parameter calculations. It was also found that water level climatology can be useful in significantly reducing the estimation uncertainty for these narrow rivers. While reach averaging length appeared insensitive to accuracy for wide rivers (wid- h ${bf gt! 1nbsphbox{km}}$), a few rivers seemed to have an optimal reach averaging length at which the highest accuracy is obtained. Finally, it was found that if reach-averaged hydraulic parameters (area, slope, and radius) are used for the calculation of reach-averaged discharge, the needed linear (bias) correction factors, although unique and arbitrary for each river reach, can improve accuracy of flow simulations.