River geometry challenges for large-scale dynamic river modeling

Traditionally, large-scale river network models in hydrology have used reduced physics, e.g. Muskingum-Cunge, to handle unsteady river flows. Recently, it has been demonstrated that efficient dynamic solutions of the 1D Saint-Venant equations can be applied over large river networks with the Simulation Program for River Networks (SPRNT). The advantage of solving the Saint-Venant equations is that models can represent water depth (as well as flow) throughout the network. However, this approach is not without challenges. In particular, a Saint-Venant solution must include representation of channel geometry, which can be difficult to consistently and comprehensively obtain over a large river network. In the present work, we explore methods for combining different data sources into a consistent river network geometric description. High-resolution lidar, cross-section surveys, and data from the National Hydrography Dataset (NHD) are merged to create reach-scale geometric descriptions of a river network suitable for Saint-Venant modeling. The Upper Alabama River is used as a study site. In this data set, 3.3% (169 out of 5069) of the reaches showed significant mismatches between available high-resolution DEMs and the NHD flowlines, which creates uncertainty as to the quality of the resulting geometric model. We investigate simplified methods for representing approximate cross-section geometry and the effect of these simplifications on flow simulations. Acknowledgement: This research was supported by the National Science Foundation under grant number CCF-1331610.